{
 "schema_version": "1.1.0",
 "benchmark_name": "PascalBench",
 "subtitle": "A legal work benchmark by Index Law",
 "brand_line": "Pascal, an experimental AI lawyer by Index Law. Technical research preview.",
 "version": "1.3",
 "last_updated": "2026-10-31T16:12:01-07:00",
 "dataset": {
  "sha256": "5a861b05eda31e3734b31265aa323e1a23a25c2d119580f176f7e341289f241d",
  "results_sha256": "c587bd8eab240382a8a071db6fc45e8116c6bb7b5d88f7961930bf6b0e352f06",
  "files": 213,
  "note": "sha256 over (path, file sha256) for task files, customer task files, rubric, routing cases, coverage and optimizer config"
 },
 "counts": {
  "tasks": 171,
  "customer_tasks": 38,
  "cases": 199,
  "cases_by_source": {
   "hill-climber": 24,
   "qa-pod": 3,
   "main-worker": 26,
   "focus-group": 44,
   "realistic": 102
  },
  "realistic_set_size": 102,
  "runs": 51,
  "models": 7,
  "model_outputs_scored": 197,
  "citations": 438,
  "rows": 1456
 },
 "models": [
  "anthropic/claude-haiku-5.5",
  "pascal-1-fallback",
  "google/gemini-4-argon",
  "openai/gpt-6-astra",
  "pascal-1",
  "openai/gpt-6.1-terra",
  "x-ai/grok-4.7"
 ],
 "pascal_model": "openai/gpt-6-astra",
 "judge": {
  "model": "pascal-1",
  "is_contestant": true,
  "scale": "integer 0 to 5 per dimension",
  "formula": "task score = (accuracy + grounding + specificity + actionability + polish + 2 x e2e_wow) / 7; hard fail caps at 2.0; overall = task-weighted mean",
  "second_judge": {
   "model": "google/gemini-4-argon",
   "n": 42,
   "pearson": 0.3085,
   "spearman": 0.2853,
   "mad": 1.7686,
   "mean_shift": 1.7686
  },
  "retest": {
   "n": 7,
   "mad": 0.06,
   "recomputed": false,
   "source": "evals/model-bakeoff.md"
  },
  "human_agreement": {
   "status": "not_yet_measured",
   "value": null,
   "n": null
  }
 },
 "rubric": {
  "source": "evals/rubric.md",
  "dimensions": [
   {
    "id": "accuracy",
    "label": "Accuracy",
    "weight": 1.0
   },
   {
    "id": "grounding",
    "label": "Grounding",
    "weight": 1.0
   },
   {
    "id": "specificity",
    "label": "Specificity",
    "weight": 1.0
   },
   {
    "id": "actionability",
    "label": "Actionability",
    "weight": 1.0
   },
   {
    "id": "polish",
    "label": "Polish",
    "weight": 1.0
   },
   {
    "id": "e2e_wow",
    "label": "End to end (x2)",
    "weight": 2.0
   }
  ],
  "hard_fails": [
   "Answers a different question than asked.",
   "Invents a statute section, rule number or case.",
   "Sends, or claims to have sent, an email without explicit approval.",
   "Drops or renames a defined term in a redline of the user's own document."
  ],
  "hard_fail_cap": 2.0
 },
 "splits": {
  "train": [
   "CG1",
   "CG3",
   "R1",
   "R3",
   "FG-L1",
   "FG-L3",
   "01",
   "02",
   "06",
   "08",
   "09",
   "10",
   "12",
   "13",
   "15",
   "20",
   "21"
  ],
  "heldout": [
   "CG2",
   "CG4",
   "R2",
   "R4",
   "FG-L2",
   "03",
   "05",
   "11",
   "18",
   "22",
   "23"
  ],
  "common6": [
   "CG1",
   "CG2",
   "CG3",
   "CG4",
   "01",
   "FG-L1"
  ],
  "rule": "FG-L4 (latency) is measured by timing, not by the judge, so it is not in either split. The critic only ever sees train outputs. Held-out tasks are scored, never shown to the critic."
 },
 "environment": {
  "status": "partial",
  "run_id": "R-CONTRACT-REALISM",
  "run_date": "2026-10-31",
  "harness_patch": "0001-contract-hillclimb",
  "card": {
   "state": "request + resolved entity (legal name, form, incorporation, number, office, signatory) + Vault context (playbook positions, files)",
   "action": "the draft, from the production drafter on the same prompt path the app uses",
   "reward": "rubric v3 task score (weights 3,3,3,2,1,1,2 on 0 to 5) capped at 1.5 on any judge hard fail, pre-check auto-fail or P0; round score 0 to 100 = half task score, half 100 minus weighted defects",
   "reward_measured": "realism rubric mean, 7 dimensions, 0 to 5, reviewer-graded",
   "adversary": "jurisdiction review panel: senior partner + associate as opposing counsel, emitting P0 to P3 findings with clause, issue, exact fix and generalised_rule",
   "transition": "repair: the drafter receives its own draft plus the findings, then the panel reviews again",
   "terminal": "0 P0, 0 P1 and fewer than 2 P2, or 4 reviews (R0 to R3); ship rule 0 P0 and 0 P1",
   "gate": "prompt patch kept only if the total score does not drop and no case gains a P0 or P1 versus baseline (greedy)",
   "panels": [
    {
     "code": "EW",
     "label": "England and Wales: senior partner (25y) + 5th-year associate"
    },
    {
     "code": "US",
     "label": "US: AmLaw 20 partner (25y) + 4th-year associate"
    },
    {
     "code": "GULF",
     "label": "Gulf: partner (20y, DIFC rights of audience) + mid-level associate"
    }
   ],
   "severities": [
    {
     "id": "P0",
     "label": "wrong in law or unusable: do not send",
     "weight": 25
    },
    {
     "id": "P1",
     "label": "a partner would send it back",
     "weight": 10
    },
    {
     "id": "P2",
     "label": "a senior associate would fix it before sending",
     "weight": 3
    },
    {
     "id": "P3",
     "label": "polish",
     "weight": 1
    }
   ]
  },
  "rounds": [
   {
    "round": 0,
    "n": 12,
    "observed": 12,
    "P0": 19,
    "P1": 33,
    "P2": 23,
    "P3": 2,
    "reward_mean": 2.3333,
    "reward_ci_low": 1.7158,
    "reward_ci_high": 2.9158,
    "defect_score_mean": 32.1667,
    "shippable": 0,
    "clean": 0
   },
   {
    "round": 1,
    "n": 12,
    "observed": 11,
    "P0": 0,
    "P1": 3,
    "P2": 7,
    "P3": 9,
    "reward_mean": 4.3325,
    "reward_ci_low": 3.8933,
    "reward_ci_high": 4.64,
    "defect_score_mean": 95.0,
    "shippable": 10,
    "clean": 9
   },
   {
    "round": 2,
    "n": 12,
    "observed": 3,
    "P0": 0,
    "P1": 2,
    "P2": 4,
    "P3": 10,
    "reward_mean": 4.4042,
    "reward_ci_low": 3.9525,
    "reward_ci_high": 4.7242,
    "defect_score_mean": 96.5,
    "shippable": 11,
    "clean": 11
   }
  ],
  "learning": {
   "delta_r0_r2": 2.0708,
   "ci_low": 1.4025,
   "ci_high": 2.7617,
   "p_sign_flip": 0.001,
   "n": 12,
   "improved": 11,
   "unchanged": 1,
   "delta_r0_r1": 1.9992,
   "test": "paired percentile bootstrap over samples (10,000 resamples) and exact sign-flip permutation (2^12), last round carried forward"
  },
  "dimensions": [
   "parties_execution",
   "structure_definitions",
   "commercial_mechanics",
   "law_disputes",
   "boilerplate",
   "legal_currency",
   "house_style"
  ],
  "dimension_rounds": [
   {
    "round": 0,
    "parties_execution": 1.75,
    "structure_definitions": 2.0833,
    "commercial_mechanics": 2.0833,
    "law_disputes": 2.75,
    "boilerplate": 2.0833,
    "legal_currency": 2.9167,
    "house_style": 2.6667
   },
   {
    "round": 1,
    "parties_execution": 4.5833,
    "structure_definitions": 4.3333,
    "commercial_mechanics": 4.0,
    "law_disputes": 4.5833,
    "boilerplate": 4.3333,
    "legal_currency": 4.6667,
    "house_style": 3.8333
   },
   {
    "round": 2,
    "parties_execution": 4.5833,
    "structure_definitions": 4.5,
    "commercial_mechanics": 4.1667,
    "law_disputes": 4.5833,
    "boilerplate": 4.4167,
    "legal_currency": 4.6667,
    "house_style": 3.9167
   }
  ],
  "episodes": [
   {
    "sample": "S01",
    "segment": "Onboarding sample",
    "doc_type": "Mutual NDA",
    "jurisdiction": "Unknown, EW default",
    "panel": "EW",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 0.86,
      "defect_score": 0,
      "P0": 3,
      "P1": 3,
      "P2": 2,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.43,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S02",
    "segment": "Signing fixture",
    "doc_type": "NDA and MFN side letter",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "users_see": false,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 0.43,
      "defect_score": 12,
      "P0": 3,
      "P1": 1,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.43,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S03",
    "segment": "Drafting engine",
    "doc_type": "NDA",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 3.43,
      "defect_score": 25,
      "P0": 1,
      "P1": 4,
      "P2": 3,
      "P3": 1,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.71,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S04",
    "segment": "Drafting engine",
    "doc_type": "NDA",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 3.14,
      "defect_score": 46,
      "P0": 1,
      "P1": 2,
      "P2": 3,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.57,
      "defect_score": 96,
      "P0": 0,
      "P1": 0,
      "P2": 1,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S05",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 3.71,
      "defect_score": 60,
      "P0": 0,
      "P1": 3,
      "P2": 3,
      "P3": 1,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.71,
      "defect_score": 96,
      "P0": 0,
      "P1": 0,
      "P2": 1,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S06",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "DIFC",
    "panel": "GULF",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 3.71,
      "defect_score": 71,
      "P0": 0,
      "P1": 2,
      "P2": 3,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.57,
      "defect_score": 96,
      "P0": 0,
      "P1": 0,
      "P2": 1,
      "P3": 1,
      "shippable": true,
      "clean": true
     },
     {
      "round": 2,
      "reward": 4.71,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1; R2 run to move a misplaced clause",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S07",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "UAE onshore",
    "panel": "GULF",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 3.14,
      "defect_score": 39,
      "P0": 1,
      "P1": 3,
      "P2": 2,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.71,
      "defect_score": 94,
      "P0": 0,
      "P1": 0,
      "P2": 2,
      "P3": 0,
      "shippable": true,
      "clean": false
     },
     {
      "round": 2,
      "reward": 4.86,
      "defect_score": 97,
      "P0": 0,
      "P1": 0,
      "P2": 1,
      "P3": 0,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 2,
    "terminal": "clean at R2 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S08",
    "segment": "Intent fixture",
    "doc_type": "Consulting agreement",
    "jurisdiction": "California",
    "panel": "US",
    "users_see": false,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 1.43,
      "defect_score": 0,
      "P0": 2,
      "P1": 5,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.71,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S09",
    "segment": "Intent fixture",
    "doc_type": "SAFE with MFN",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": false,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 1.86,
      "defect_score": 32,
      "P0": 1,
      "P1": 4,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.29,
      "defect_score": 87,
      "P0": 0,
      "P1": 1,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 2,
      "reward": 4.86,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 2,
    "terminal": "clean at R2 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S10",
    "segment": "Intent fixture",
    "doc_type": "SAFE, post-money template",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": false,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 1.71,
      "defect_score": 24,
      "P0": 2,
      "P1": 2,
      "P2": 2,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 4.86,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S11",
    "segment": "System prompt",
    "doc_type": "Jurisdiction law blocks",
    "jurisdiction": "UAE, DIFC, ADGM, England",
    "panel": "GULF",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 2.29,
      "defect_score": 0,
      "P0": 5,
      "P1": 2,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     },
     {
      "round": 1,
      "reward": 3.71,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 1,
    "terminal": "clean at R1 (stop rule)",
    "shippable_r0": false,
    "shippable_last": true
   },
   {
    "sample": "S12",
    "segment": "Demo seed prompt",
    "doc_type": "Demo contract set",
    "jurisdiction": "US parent, UK subsidiary",
    "panel": "US",
    "users_see": true,
    "on_main": true,
    "rounds": [
     {
      "round": 0,
      "reward": 2.29,
      "defect_score": 77,
      "P0": 0,
      "P1": 2,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (outputs not readable offline)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-consent_letter",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Consent letter",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 2.43,
      "defect_score": 12,
      "P0": 3,
      "P1": 1,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-mutual_nda",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Mutual NDA",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 3.71,
      "defect_score": 64,
      "P0": 0,
      "P1": 3,
      "P2": 2,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-safe",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "SAFE",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 4.71,
      "defect_score": 87,
      "P0": 0,
      "P1": 1,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-mfn_waiver",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "MFN waiver",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 4.71,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 0,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": true,
    "shippable_last": true
   },
   {
    "sample": "D-side_letter",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Side letter",
    "jurisdiction": "New York",
    "panel": "US",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 2.71,
      "defect_score": 77,
      "P0": 0,
      "P1": 2,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-difc_services",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Services agreement",
    "jurisdiction": "DIFC",
    "panel": "GULF",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 4.43,
      "defect_score": 87,
      "P0": 0,
      "P1": 1,
      "P2": 1,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-uae_services",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Services agreement",
    "jurisdiction": "UAE onshore",
    "panel": "GULF",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 4.43,
      "defect_score": 80,
      "P0": 0,
      "P1": 2,
      "P2": 0,
      "P3": 0,
      "shippable": false,
      "clean": false
     }
    ],
    "rounds_to_clean": null,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": false,
    "shippable_last": false
   },
   {
    "sample": "D-board_consent",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Board consent",
    "jurisdiction": "Delaware",
    "panel": "US",
    "users_see": false,
    "on_main": false,
    "rounds": [
     {
      "round": 0,
      "reward": 4.86,
      "defect_score": 99,
      "P0": 0,
      "P1": 0,
      "P2": 0,
      "P3": 1,
      "shippable": true,
      "clean": true
     }
    ],
    "rounds_to_clean": 0,
    "terminal": "not repaired (drafting-branch fixture, R0 only)",
    "shippable_r0": true,
    "shippable_last": true
   }
  ],
  "explorer": [
   {
    "sample": "S01",
    "segment": "Onboarding sample",
    "doc_type": "Mutual NDA",
    "jurisdiction": "Unknown, EW default",
    "panel": "EW",
    "on_main": true,
    "round": 0,
    "reward": 0.86,
    "defect_score": 0,
    "P0": 3,
    "P1": 3,
    "P2": 2,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S01",
    "segment": "Onboarding sample",
    "doc_type": "Mutual NDA",
    "jurisdiction": "Unknown, EW default",
    "panel": "EW",
    "on_main": true,
    "round": 1,
    "reward": 4.43,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S02",
    "segment": "Signing fixture",
    "doc_type": "NDA and MFN side letter",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": true,
    "round": 0,
    "reward": 0.43,
    "defect_score": 12,
    "P0": 3,
    "P1": 1,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S02",
    "segment": "Signing fixture",
    "doc_type": "NDA and MFN side letter",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": true,
    "round": 1,
    "reward": 4.43,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S03",
    "segment": "Drafting engine",
    "doc_type": "NDA",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": true,
    "round": 0,
    "reward": 3.43,
    "defect_score": 25,
    "P0": 1,
    "P1": 4,
    "P2": 3,
    "P3": 1,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S03",
    "segment": "Drafting engine",
    "doc_type": "NDA",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": true,
    "round": 1,
    "reward": 4.71,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S04",
    "segment": "Drafting engine",
    "doc_type": "NDA",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 0,
    "reward": 3.14,
    "defect_score": 46,
    "P0": 1,
    "P1": 2,
    "P2": 3,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S04",
    "segment": "Drafting engine",
    "doc_type": "NDA",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 1,
    "reward": 4.57,
    "defect_score": 96,
    "P0": 0,
    "P1": 0,
    "P2": 1,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S05",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": true,
    "round": 0,
    "reward": 3.71,
    "defect_score": 60,
    "P0": 0,
    "P1": 3,
    "P2": 3,
    "P3": 1,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S05",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": true,
    "round": 1,
    "reward": 4.71,
    "defect_score": 96,
    "P0": 0,
    "P1": 0,
    "P2": 1,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S06",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "DIFC",
    "panel": "GULF",
    "on_main": true,
    "round": 0,
    "reward": 3.71,
    "defect_score": 71,
    "P0": 0,
    "P1": 2,
    "P2": 3,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S06",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "DIFC",
    "panel": "GULF",
    "on_main": true,
    "round": 1,
    "reward": 4.57,
    "defect_score": 96,
    "P0": 0,
    "P1": 0,
    "P2": 1,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S06",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "DIFC",
    "panel": "GULF",
    "on_main": true,
    "round": 2,
    "reward": 4.71,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S07",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "UAE onshore",
    "panel": "GULF",
    "on_main": true,
    "round": 0,
    "reward": 3.14,
    "defect_score": 39,
    "P0": 1,
    "P1": 3,
    "P2": 2,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S07",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "UAE onshore",
    "panel": "GULF",
    "on_main": true,
    "round": 1,
    "reward": 4.71,
    "defect_score": 94,
    "P0": 0,
    "P1": 0,
    "P2": 2,
    "P3": 0,
    "shippable": true,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S07",
    "segment": "Drafting engine",
    "doc_type": "Services agreement",
    "jurisdiction": "UAE onshore",
    "panel": "GULF",
    "on_main": true,
    "round": 2,
    "reward": 4.86,
    "defect_score": 97,
    "P0": 0,
    "P1": 0,
    "P2": 1,
    "P3": 0,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S08",
    "segment": "Intent fixture",
    "doc_type": "Consulting agreement",
    "jurisdiction": "California",
    "panel": "US",
    "on_main": true,
    "round": 0,
    "reward": 1.43,
    "defect_score": 0,
    "P0": 2,
    "P1": 5,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S08",
    "segment": "Intent fixture",
    "doc_type": "Consulting agreement",
    "jurisdiction": "California",
    "panel": "US",
    "on_main": true,
    "round": 1,
    "reward": 4.71,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S09",
    "segment": "Intent fixture",
    "doc_type": "SAFE with MFN",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 0,
    "reward": 1.86,
    "defect_score": 32,
    "P0": 1,
    "P1": 4,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S09",
    "segment": "Intent fixture",
    "doc_type": "SAFE with MFN",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 1,
    "reward": 4.29,
    "defect_score": 87,
    "P0": 0,
    "P1": 1,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S09",
    "segment": "Intent fixture",
    "doc_type": "SAFE with MFN",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 2,
    "reward": 4.86,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S10",
    "segment": "Intent fixture",
    "doc_type": "SAFE, post-money template",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 0,
    "reward": 1.71,
    "defect_score": 24,
    "P0": 2,
    "P1": 2,
    "P2": 2,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S10",
    "segment": "Intent fixture",
    "doc_type": "SAFE, post-money template",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": true,
    "round": 1,
    "reward": 4.86,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S11",
    "segment": "System prompt",
    "doc_type": "Jurisdiction law blocks",
    "jurisdiction": "UAE, DIFC, ADGM, England",
    "panel": "GULF",
    "on_main": true,
    "round": 0,
    "reward": 2.29,
    "defect_score": 0,
    "P0": 5,
    "P1": 2,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S11",
    "segment": "System prompt",
    "doc_type": "Jurisdiction law blocks",
    "jurisdiction": "UAE, DIFC, ADGM, England",
    "panel": "GULF",
    "on_main": true,
    "round": 1,
    "reward": 3.71,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "S12",
    "segment": "Demo seed prompt",
    "doc_type": "Demo contract set",
    "jurisdiction": "US parent, UK subsidiary",
    "panel": "US",
    "on_main": true,
    "round": 0,
    "reward": 2.29,
    "defect_score": 77,
    "P0": 0,
    "P1": 2,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-consent_letter",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Consent letter",
    "jurisdiction": "England and Wales",
    "panel": "EW",
    "on_main": false,
    "round": 0,
    "reward": 2.43,
    "defect_score": 12,
    "P0": 3,
    "P1": 1,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-mutual_nda",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Mutual NDA",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": false,
    "round": 0,
    "reward": 3.71,
    "defect_score": 64,
    "P0": 0,
    "P1": 3,
    "P2": 2,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-safe",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "SAFE",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": false,
    "round": 0,
    "reward": 4.71,
    "defect_score": 87,
    "P0": 0,
    "P1": 1,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-mfn_waiver",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "MFN waiver",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": false,
    "round": 0,
    "reward": 4.71,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-side_letter",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Side letter",
    "jurisdiction": "New York",
    "panel": "US",
    "on_main": false,
    "round": 0,
    "reward": 2.71,
    "defect_score": 77,
    "P0": 0,
    "P1": 2,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-difc_services",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Services agreement",
    "jurisdiction": "DIFC",
    "panel": "GULF",
    "on_main": false,
    "round": 0,
    "reward": 4.43,
    "defect_score": 87,
    "P0": 0,
    "P1": 1,
    "P2": 1,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-uae_services",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Services agreement",
    "jurisdiction": "UAE onshore",
    "panel": "GULF",
    "on_main": false,
    "round": 0,
    "reward": 4.43,
    "defect_score": 80,
    "P0": 0,
    "P1": 2,
    "P2": 0,
    "P3": 0,
    "shippable": false,
    "clean": false,
    "basis": "reviewer_graded"
   },
   {
    "sample": "D-board_consent",
    "segment": "ContractDoc fixture (branch)",
    "doc_type": "Board consent",
    "jurisdiction": "Delaware",
    "panel": "US",
    "on_main": false,
    "round": 0,
    "reward": 4.86,
    "defect_score": 99,
    "P0": 0,
    "P1": 0,
    "P2": 0,
    "P3": 1,
    "shippable": true,
    "clean": true,
    "basis": "reviewer_graded"
   }
  ],
  "heatmap": {
   "categories": [
    "unenforceable",
    "wrong_jurisdiction",
    "missing_mandatory_term",
    "defined_term_drift",
    "placeholder",
    "commercial_nonsense",
    "ai_tell",
    "house_style",
    "personalisation",
    "stale_law"
   ],
   "columns": [
    "R0",
    "R1",
    "R2",
    "Branch R0"
   ],
   "cells": [
    {
     "category": "unenforceable",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 3,
     "P0": 0,
     "P1": 2,
     "P2": 1,
     "P3": 0
    },
    {
     "category": "unenforceable",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 1,
     "P0": 0,
     "P1": 0,
     "P2": 1,
     "P3": 0
    },
    {
     "category": "unenforceable",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 1,
     "P0": 0,
     "P1": 0,
     "P2": 1,
     "P3": 0
    },
    {
     "category": "unenforceable",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 1,
     "P0": 1,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "wrong_jurisdiction",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 12,
     "P0": 2,
     "P1": 3,
     "P2": 7,
     "P3": 0
    },
    {
     "category": "wrong_jurisdiction",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 2,
     "P0": 0,
     "P1": 0,
     "P2": 1,
     "P3": 1
    },
    {
     "category": "wrong_jurisdiction",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 1,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 1
    },
    {
     "category": "wrong_jurisdiction",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 2,
     "P0": 0,
     "P1": 0,
     "P2": 2,
     "P3": 0
    },
    {
     "category": "missing_mandatory_term",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 28,
     "P0": 6,
     "P1": 17,
     "P2": 5,
     "P3": 0
    },
    {
     "category": "missing_mandatory_term",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 2,
     "P0": 0,
     "P1": 0,
     "P2": 1,
     "P3": 1
    },
    {
     "category": "missing_mandatory_term",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "missing_mandatory_term",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 13,
     "P0": 1,
     "P1": 8,
     "P2": 3,
     "P3": 1
    },
    {
     "category": "defined_term_drift",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 5,
     "P0": 1,
     "P1": 2,
     "P2": 2,
     "P3": 0
    },
    {
     "category": "defined_term_drift",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 3,
     "P0": 0,
     "P1": 1,
     "P2": 1,
     "P3": 1
    },
    {
     "category": "defined_term_drift",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 1,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 1
    },
    {
     "category": "defined_term_drift",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 3,
     "P0": 0,
     "P1": 1,
     "P2": 1,
     "P3": 1
    },
    {
     "category": "placeholder",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 5,
     "P0": 3,
     "P1": 1,
     "P2": 1,
     "P3": 0
    },
    {
     "category": "placeholder",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "placeholder",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "placeholder",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 1,
     "P0": 1,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "commercial_nonsense",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 7,
     "P0": 2,
     "P1": 5,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "commercial_nonsense",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "commercial_nonsense",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "commercial_nonsense",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "ai_tell",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "ai_tell",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "ai_tell",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "ai_tell",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "house_style",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 9,
     "P0": 0,
     "P1": 0,
     "P2": 7,
     "P3": 2
    },
    {
     "category": "house_style",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 5,
     "P0": 0,
     "P1": 0,
     "P2": 2,
     "P3": 3
    },
    {
     "category": "house_style",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "house_style",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 1,
     "P0": 0,
     "P1": 1,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "personalisation",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 2,
     "P0": 0,
     "P1": 2,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "personalisation",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 2,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 2
    },
    {
     "category": "personalisation",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "personalisation",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "stale_law",
     "column": "R0",
     "round": 0,
     "on_main": true,
     "n": 6,
     "P0": 5,
     "P1": 1,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "stale_law",
     "column": "R1",
     "round": 1,
     "on_main": true,
     "n": 1,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 1
    },
    {
     "category": "stale_law",
     "column": "R2",
     "round": 2,
     "on_main": true,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    },
    {
     "category": "stale_law",
     "column": "Branch R0",
     "round": 0,
     "on_main": false,
     "n": 0,
     "P0": 0,
     "P1": 0,
     "P2": 0,
     "P3": 0
    }
   ],
   "method": "ordered keyword rules over issue and generalised_rule (environment.py CAT_RULES)"
  },
  "policy_updates": [
   {
    "rank": 1,
    "rule": "Parties block completeness is a precondition, not an optional field.",
    "category": "missing_mandatory_term",
    "findings": 8,
    "weight": 80,
    "max_severity": "P1",
    "P0": 0,
    "P1": 8,
    "P2": 0,
    "P3": 0,
    "samples": [
     "D-consent_letter",
     "D-difc_services",
     "D-uae_services",
     "S03",
     "S06",
     "S07",
     "S08",
     "S09"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-01",
     "P-02",
     "P-13",
     "P-14",
     "P-15"
    ]
   },
   {
    "rank": 2,
    "rule": "Prompts cite current law only; a law-change table drives both prompts and engine.",
    "category": "stale_law",
    "findings": 3,
    "weight": 75,
    "max_severity": "P0",
    "P0": 3,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S11"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-12",
     "P-15"
    ]
   },
   {
    "rank": 3,
    "rule": "Every contract sample has a governing law and a jurisdiction-correct forum clause.",
    "category": "missing_mandatory_term",
    "findings": 2,
    "weight": 50,
    "max_severity": "P0",
    "P0": 2,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S01",
     "S02"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-09",
     "P-10",
     "P-15"
    ]
   },
   {
    "rank": 4,
    "rule": "Seeds and fixtures are realistic documents, not mail-merge shells.",
    "category": "placeholder",
    "findings": 2,
    "weight": 50,
    "max_severity": "P0",
    "P0": 2,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S08",
     "S10"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-13",
     "P-15"
    ]
   },
   {
    "rank": 5,
    "rule": "A document must contain the operative obligation its title promises; an NDA without a confidentiality covenant fails the eval.",
    "category": "missing_mandatory_term",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S01"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-09"
    ]
   },
   {
    "rank": 6,
    "rule": "A sample presented as signed is complete.",
    "category": "missing_mandatory_term",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S09"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-13"
    ]
   },
   {
    "rank": 7,
    "rule": "Entity form must match the legal suffix and the register; never fall back to 'a corporation'.",
    "category": "wrong_jurisdiction",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S04"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-01"
    ]
   },
   {
    "rank": 8,
    "rule": "Every cross-referenced schedule exists.",
    "category": "commercial_nonsense",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S08"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-13"
    ]
   },
   {
    "rank": 9,
    "rule": "Every economic term resolves to defined inputs.",
    "category": "commercial_nonsense",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S10"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-13"
    ]
   },
   {
    "rank": 10,
    "rule": "Every party is described with full legal name, entity type, jurisdiction of incorporation, company number where customary and registered or principal office.",
    "category": "missing_mandatory_term",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S01"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-09",
     "P-15"
    ]
   },
   {
    "rank": 11,
    "rule": "Examples in prompts are current and verified; models copy them.",
    "category": "stale_law",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S11"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-12",
     "P-15"
    ]
   },
   {
    "rank": 12,
    "rule": "Fixtures that render a signable document reuse a vetted sample constant; no inline stubs.",
    "category": "placeholder",
    "findings": 1,
    "weight": 25,
    "max_severity": "P0",
    "P0": 1,
    "P1": 0,
    "P2": 0,
    "P3": 0,
    "samples": [
     "S02"
    ],
    "rounds": [
     0
    ],
    "patches": [
     "P-10",
     "P-15"
    ]
   }
  ],
  "policy_clusters": 89,
  "patches": [
   {
    "id": "P-01",
    "title": "Never guess entity type or place of incorporation",
    "owner": "DL"
   },
   {
    "id": "P-02",
    "title": "Notices point at real addresses, with business-hours receipt",
    "owner": "DL"
   },
   {
    "id": "P-03",
    "title": "Governing-law-aware boilerplate and signature blocks",
    "owner": "DL"
   },
   {
    "id": "P-04",
    "title": "NDA short names, Affiliate, Representatives, attestation",
    "owner": "DL"
   },
   {
    "id": "P-05",
    "title": "Services: computable cap, fee carve-out, Art 28(3) particulars, pronouns, further assurance",
    "owner": "DL"
   },
   {
    "id": "P-06",
    "title": "No services agreement without services; regulatory perimeter in DIFC and ADGM",
    "owner": "DL"
   },
   {
    "id": "P-07",
    "title": "Onshore UAE arbitration: signatory authority",
    "owner": "DL"
   },
   {
    "id": "P-08",
    "title": "Self-check accepts plural uses of a defined term",
    "owner": "DL"
   },
   {
    "id": "P-09",
    "title": "Replace the onboarding sample with a complete English-law mutual NDA",
    "owner": "ON"
   },
   {
    "id": "P-10",
    "title": "Signing fixtures use vetted sample text",
    "owner": "SG"
   },
   {
    "id": "P-11",
    "title": "Realism bar for every prompt that drafts or seeds contracts",
    "owner": "DL"
   },
   {
    "id": "P-12",
    "title": "Current law in the jurisdiction blocks and the citation example",
    "owner": "PR"
   },
   {
    "id": "P-13",
    "title": "Fixtures are signed precedents, not mail-merge shells",
    "owner": "EV"
   },
   {
    "id": "P-14",
    "title": "ContractDoc fixtures on feat/drafting-contractdoc",
    "owner": "DL"
   },
   {
    "id": "P-15",
    "title": "Lint in CI",
    "owner": "EV"
   }
  ],
  "precheck": {
   "n": 221,
   "auto_fail": 141,
   "clean": 80,
   "cap": 1.5,
   "by_reason": [
    {
     "reason": "over_budget",
     "outputs": 122,
     "only_reason": 112
    },
    {
     "reason": "no_citations",
     "outputs": 29,
     "only_reason": 17
    },
    {
     "reason": "em_dash",
     "outputs": 2,
     "only_reason": 0
    },
    {
     "reason": "placeholder",
     "outputs": 1,
     "only_reason": 0
    },
    {
     "reason": "party_mismatch",
     "outputs": 0,
     "only_reason": 0
    }
   ],
   "by_model": [
    {
     "key": "google/gemini-4-argon",
     "n": 10,
     "auto_fail": 10,
     "rate": 1.0,
     "ci_low": 0.7225,
     "ci_high": 1.0,
     "clean": 0,
     "over_budget": 10,
     "no_citations": 0,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "not recorded",
     "n": 19,
     "auto_fail": 18,
     "rate": 0.9474,
     "ci_low": 0.7536,
     "ci_high": 0.9906,
     "clean": 1,
     "over_budget": 15,
     "no_citations": 9,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "x-ai/grok-4.7",
     "n": 10,
     "auto_fail": 9,
     "rate": 0.9,
     "ci_low": 0.5958,
     "ci_high": 0.9821,
     "clean": 1,
     "over_budget": 8,
     "no_citations": 1,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "openai/gpt-6-astra",
     "n": 12,
     "auto_fail": 10,
     "rate": 0.8333,
     "ci_low": 0.552,
     "ci_high": 0.953,
     "clean": 2,
     "over_budget": 7,
     "no_citations": 4,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "anthropic/claude-haiku-5.5",
     "n": 10,
     "auto_fail": 8,
     "rate": 0.8,
     "ci_low": 0.4902,
     "ci_high": 0.9433,
     "clean": 2,
     "over_budget": 6,
     "no_citations": 3,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "pascal-1-fallback",
     "n": 10,
     "auto_fail": 7,
     "rate": 0.7,
     "ci_low": 0.3968,
     "ci_high": 0.8922,
     "clean": 3,
     "over_budget": 7,
     "no_citations": 1,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "pascal-1",
     "n": 140,
     "auto_fail": 74,
     "rate": 0.5286,
     "ci_low": 0.4462,
     "ci_high": 0.6094,
     "clean": 66,
     "over_budget": 65,
     "no_citations": 9,
     "em_dash": 0,
     "placeholder": 1,
     "party_mismatch": 0
    },
    {
     "key": "openai/gpt-6.1-terra",
     "n": 10,
     "auto_fail": 5,
     "rate": 0.5,
     "ci_low": 0.2366,
     "ci_high": 0.7634,
     "clean": 5,
     "over_budget": 4,
     "no_citations": 2,
     "em_dash": 2,
     "placeholder": 0,
     "party_mismatch": 0
    }
   ],
   "by_prompt": [
    {
     "key": "production app",
     "n": 19,
     "auto_fail": 18,
     "rate": 0.9474,
     "ci_low": 0.7536,
     "ci_high": 0.9906,
     "clean": 1,
     "over_budget": 15,
     "no_citations": 9,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "optimizer v1 (pack v2.6 s0-7)",
     "n": 5,
     "auto_fail": 4,
     "rate": 0.8,
     "ci_low": 0.3755,
     "ci_high": 0.9638,
     "clean": 1,
     "over_budget": 3,
     "no_citations": 1,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "production QUALITY_BAR",
     "n": 17,
     "auto_fail": 13,
     "rate": 0.7647,
     "ci_low": 0.5274,
     "ci_high": 0.9045,
     "clean": 4,
     "over_budget": 5,
     "no_citations": 9,
     "em_dash": 0,
     "placeholder": 1,
     "party_mismatch": 0
    },
    {
     "key": "pack v2.6",
     "n": 66,
     "auto_fail": 50,
     "rate": 0.7576,
     "ci_low": 0.6419,
     "ci_high": 0.8449,
     "clean": 16,
     "over_budget": 46,
     "no_citations": 7,
     "em_dash": 2,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "optimizer v1",
     "n": 38,
     "auto_fail": 20,
     "rate": 0.5263,
     "ci_low": 0.3726,
     "ci_high": 0.6752,
     "clean": 18,
     "over_budget": 20,
     "no_citations": 0,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "optimizer v2",
     "n": 38,
     "auto_fail": 20,
     "rate": 0.5263,
     "ci_low": 0.3726,
     "ci_high": 0.6752,
     "clean": 18,
     "over_budget": 18,
     "no_citations": 2,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "optimizer v3",
     "n": 19,
     "auto_fail": 8,
     "rate": 0.4211,
     "ci_low": 0.2314,
     "ci_high": 0.6372,
     "clean": 11,
     "over_budget": 7,
     "no_citations": 1,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    },
    {
     "key": "optimizer v4",
     "n": 19,
     "auto_fail": 8,
     "rate": 0.4211,
     "ci_low": 0.2314,
     "ci_high": 0.6372,
     "clean": 11,
     "over_budget": 8,
     "no_citations": 0,
     "em_dash": 0,
     "placeholder": 0,
     "party_mismatch": 0
    }
   ],
   "budget_sensitivity": [
    {
     "multiplier": 1.0,
     "auto_fail": 141,
     "n": 221
    },
    {
     "multiplier": 1.25,
     "auto_fail": 94,
     "n": 221
    },
    {
     "multiplier": 1.5,
     "auto_fail": 59,
     "n": 221
    },
    {
     "multiplier": 2.0,
     "auto_fail": 37,
     "n": 221
    }
   ],
   "without_length_check": 29
  },
  "harness": {
   "cases": 17,
   "fixtures": 6,
   "fixtures_ok": 6,
   "static_checks": "pass",
   "cases_by_panel": [
    {
     "panel": "Delaware and California",
     "n": 2
    },
    {
     "panel": "Delaware and New York",
     "n": 3
    },
    {
     "panel": "DIFC and ADGM",
     "n": 4
    },
    {
     "panel": "England and Wales",
     "n": 5
    },
    {
     "panel": "EU (fourth panel)",
     "n": 3
    }
   ],
   "live_run": "not_yet_measured",
   "propose_patches": "not_yet_measured",
   "mode": "static dry run, no model calls"
  },
  "optimizer_gate": [
   {
    "version": "v1",
    "parent": null,
    "heldout": 3.51,
    "train": 3.28,
    "heldout_hard_fails": 0,
    "accepted": true,
    "reason": "baseline"
   },
   {
    "version": "v2",
    "parent": "v1",
    "heldout": 3.444,
    "train": 3.299,
    "heldout_hard_fails": 0,
    "accepted": false,
    "reason": "heldout 3.444 < 3.51 + 0.05"
   },
   {
    "version": "v3",
    "parent": "v1",
    "heldout": 3.562,
    "train": 3.44,
    "heldout_hard_fails": 0,
    "accepted": true,
    "reason": "held-out gain, no new hard fails, train held"
   },
   {
    "version": "v4",
    "parent": "v3",
    "heldout": 3.645,
    "train": 3.451,
    "heldout_hard_fails": 0,
    "accepted": true,
    "reason": "held-out gain, no new hard fails, train held"
   }
  ],
  "measurement": [
   {
    "item": "Panel findings P0 to P3, R0 to R2",
    "basis": "reviewer_graded",
    "value": "117 findings",
    "note": "An offline model reviewer played each two-lawyer panel. It is not a lawyer and not Pascal's runtime judge."
   },
   {
    "item": "Realism rubric, 7 dimensions, 0 to 5 (reward used here)",
    "basis": "reviewer_graded",
    "value": "2.33 to 4.40",
    "note": "Unweighted mean of 7 integer scores per sample and round."
   },
   {
    "item": "Severity counts, ship and stop flags, defect score",
    "basis": "derived",
    "value": "P0 19 to 0",
    "note": "Deterministic from the reviewer's findings. Defect score = 100 minus P0 x25, P1 x10, P2 x3, P3 x1."
   },
   {
    "item": "Finding categories",
    "basis": "derived",
    "value": "10 categories",
    "note": "Ordered keyword rules in environment.py; the run predates the category field."
   },
   {
    "item": "Patch links",
    "basis": "derived",
    "value": "15 patches",
    "note": "Parsed from the Prevents lines of the patch list."
   },
   {
    "item": "Rubric v3 deterministic pre-check",
    "basis": "measured",
    "value": "141 of 221 auto-fail",
    "note": "Saved outputs, no judge."
   },
   {
    "item": "Harness static dry run",
    "basis": "measured",
    "value": "6 of 6 lint fixtures match, 17 cases assemble",
    "note": "No model calls."
   },
   {
    "item": "Prompt optimizer gate (v1 to v4)",
    "basis": "measured",
    "value": "3 of 4 versions kept",
    "note": "Held-out gate on the v2 rubric, judge pascal-1 (chart 7). Not the contract gate."
   },
   {
    "item": "Rubric v3 judge reward on contracts",
    "basis": "not_yet_measured",
    "value": null,
    "note": "Needs the live hill-climb run."
   },
   {
    "item": "Fresh R0 drafts from the production drafter (17 cases)",
    "basis": "not_yet_measured",
    "value": null,
    "note": "R0 here is the shipped artefact, not a policy sample."
   },
   {
    "item": "Contract prompt-patch regression gate (propose-patches mode)",
    "basis": "not_yet_measured",
    "value": null,
    "note": "Specified and unit tested in the 0001 patch."
   },
   {
    "item": "Panel agreement with practising lawyers",
    "basis": "not_yet_measured",
    "value": null,
    "note": "No lawyer labels yet."
   },
   {
    "item": "Repeat episodes per case (variance, pass@k)",
    "basis": "not_yet_measured",
    "value": null,
    "note": "One episode per sample."
   }
  ],
  "limitations": [
   "The review panel and the repairer were the same offline model reviewer. R1 and R2 rewards measure self-repair against its own critique, so they are upper bounds.",
   "R0 is what ships in the repo (fixtures, prompts and deterministic engine renders), not fresh samples from the drafting policy.",
   "n is 12 episodes on main, one episode per sample. The CI is over samples. It does not measure sampling variance.",
   "The reward shown is the 7-dimension realism rubric, not rubric v3. Rubric v3 scores with the 1.5 cap are not yet measured on contracts.",
   "Stopped episodes carry their last round forward, as in the source report. S12 stays at R0 because it could not be repaired offline.",
   "Finding categories and patch links are derived by deterministic rules from text, not labelled by the panel.",
   "Engine repairs were rendered from a local, uncommitted patched copy; none of patches P-01 to P-15 is applied on main."
  ],
  "charts": [
   "environment_loop",
   "environment_learning_curve",
   "environment_episodes",
   "environment_findings_heatmap",
   "environment_policy_updates",
   "environment_reward_gate"
  ],
  "source": [
   "evals/contract-realism.md",
   "evals/contract-realism/findings/summary.json",
   "evals/contract-realism/findings/round-0.json",
   "evals/contract-realism/findings/round-1.json",
   "evals/contract-realism/findings/round-2.json",
   "evals/contract-realism/findings_data.py",
   "evals/proposed/contract-realism-patches.md",
   "evals/proposed/contract-hillclimb/loop.md",
   "evals/proposed/contract-hillclimb/dry-run-output.txt",
   "evals/runs/2026-10-31-v3-precheck/results.csv",
   "evals/optimizer/versions.csv"
  ]
 },
 "results": [
  {
   "id": "score6:gpt-6-astra",
   "label": "gpt-6-astra, weighted score, 6 common tasks",
   "value": 3.3493,
   "unit": "score_0_5",
   "ci_low": 3.0873,
   "ci_high": 3.6609,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "Pascal default drafting model"
  },
  {
   "id": "cost:gpt-6-astra",
   "label": "gpt-6-astra, generation cost per task",
   "value": 0.2483,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:gpt-6-astra",
   "label": "gpt-6-astra, latency median",
   "value": 93.42,
   "unit": "seconds",
   "ci_low": 89.4456,
   "ci_high": 97.181,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:gpt-6-astra",
   "label": "gpt-6-astra, hard fails",
   "value": 0.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:gpt-6-astra",
   "label": "gpt-6-astra, accuracy",
   "value": 3.3333,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:gpt-6-astra",
   "label": "gpt-6-astra, grounding",
   "value": 4.0,
   "unit": "score_0_5",
   "ci_low": 4.0,
   "ci_high": 4.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:gpt-6-astra",
   "label": "gpt-6-astra, specificity",
   "value": 3.6667,
   "unit": "score_0_5",
   "ci_low": 3.1667,
   "ci_high": 4.3333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:gpt-6-astra",
   "label": "gpt-6-astra, actionability",
   "value": 3.6667,
   "unit": "score_0_5",
   "ci_low": 3.3333,
   "ci_high": 4.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:gpt-6-astra",
   "label": "gpt-6-astra, polish",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:gpt-6-astra",
   "label": "gpt-6-astra, end to end (x2)",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:pascal-1-fallback",
   "label": "pascal-1-fallback, weighted score, 6 common tasks",
   "value": 3.1906,
   "unit": "score_0_5",
   "ci_low": 3.0204,
   "ci_high": 3.3518,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score_all:pascal-1-fallback",
   "label": "pascal-1-fallback, weighted score, all 10 tasks",
   "value": 3.2064,
   "unit": "score_0_5",
   "ci_low": 3.062,
   "ci_high": 3.3751,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "cost:pascal-1-fallback",
   "label": "pascal-1-fallback, generation cost per task",
   "value": 0.1591,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:pascal-1-fallback",
   "label": "pascal-1-fallback, latency median",
   "value": 61.9087,
   "unit": "seconds",
   "ci_low": 55.4838,
   "ci_high": 68.1381,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:pascal-1-fallback",
   "label": "pascal-1-fallback, hard fails",
   "value": 0.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:pascal-1-fallback",
   "label": "pascal-1-fallback, accuracy",
   "value": 3.3333,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:pascal-1-fallback",
   "label": "pascal-1-fallback, grounding",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:pascal-1-fallback",
   "label": "pascal-1-fallback, specificity",
   "value": 3.5,
   "unit": "score_0_5",
   "ci_low": 3.1667,
   "ci_high": 3.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:pascal-1-fallback",
   "label": "pascal-1-fallback, actionability",
   "value": 3.6667,
   "unit": "score_0_5",
   "ci_low": 3.3333,
   "ci_high": 4.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:pascal-1-fallback",
   "label": "pascal-1-fallback, polish",
   "value": 3.3333,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:pascal-1-fallback",
   "label": "pascal-1-fallback, end to end (x2)",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:pascal-1",
   "label": "pascal-1, weighted score, 6 common tasks",
   "value": 2.9762,
   "unit": "score_0_5",
   "ci_low": 2.4215,
   "ci_high": 3.694,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score_all:pascal-1",
   "label": "pascal-1, weighted score, all 10 tasks",
   "value": 3.3386,
   "unit": "score_0_5",
   "ci_low": 2.8163,
   "ci_high": 3.9689,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "cost:pascal-1",
   "label": "pascal-1, generation cost per task",
   "value": 0.043,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:pascal-1",
   "label": "pascal-1, latency median",
   "value": 66.2052,
   "unit": "seconds",
   "ci_low": 64.1051,
   "ci_high": 68.0281,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:pascal-1",
   "label": "pascal-1, hard fails",
   "value": 1.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:pascal-1",
   "label": "pascal-1, accuracy",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:pascal-1",
   "label": "pascal-1, grounding",
   "value": 3.6667,
   "unit": "score_0_5",
   "ci_low": 2.8333,
   "ci_high": 4.3333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:pascal-1",
   "label": "pascal-1, specificity",
   "value": 3.5,
   "unit": "score_0_5",
   "ci_low": 2.8333,
   "ci_high": 4.3333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:pascal-1",
   "label": "pascal-1, actionability",
   "value": 3.6667,
   "unit": "score_0_5",
   "ci_low": 3.1667,
   "ci_high": 4.3333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:pascal-1",
   "label": "pascal-1, polish",
   "value": 3.1667,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:pascal-1",
   "label": "pascal-1, end to end (x2)",
   "value": 2.8333,
   "unit": "score_0_5",
   "ci_low": 2.1667,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:gpt-6.1-terra",
   "label": "gpt-6.1-terra, weighted score, 6 common tasks",
   "value": 2.8493,
   "unit": "score_0_5",
   "ci_low": 2.593,
   "ci_high": 3.2002,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score_all:gpt-6.1-terra",
   "label": "gpt-6.1-terra, weighted score, all 10 tasks",
   "value": 3.0794,
   "unit": "score_0_5",
   "ci_low": 2.7881,
   "ci_high": 3.4396,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "cost:gpt-6.1-terra",
   "label": "gpt-6.1-terra, generation cost per task",
   "value": 0.0182,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:gpt-6.1-terra",
   "label": "gpt-6.1-terra, latency median",
   "value": 37.4375,
   "unit": "seconds",
   "ci_low": 29.9232,
   "ci_high": 39.9488,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:gpt-6.1-terra",
   "label": "gpt-6.1-terra, hard fails",
   "value": 0.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:gpt-6.1-terra",
   "label": "gpt-6.1-terra, accuracy",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:gpt-6.1-terra",
   "label": "gpt-6.1-terra, grounding",
   "value": 3.3333,
   "unit": "score_0_5",
   "ci_low": 2.6667,
   "ci_high": 3.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:gpt-6.1-terra",
   "label": "gpt-6.1-terra, specificity",
   "value": 3.1667,
   "unit": "score_0_5",
   "ci_low": 2.6667,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:gpt-6.1-terra",
   "label": "gpt-6.1-terra, actionability",
   "value": 3.1667,
   "unit": "score_0_5",
   "ci_low": 2.6667,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:gpt-6.1-terra",
   "label": "gpt-6.1-terra, polish",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:gpt-6.1-terra",
   "label": "gpt-6.1-terra, end to end (x2)",
   "value": 2.5,
   "unit": "score_0_5",
   "ci_low": 2.1667,
   "ci_high": 2.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:grok-4.7",
   "label": "grok-4.7, weighted score, 6 common tasks",
   "value": 2.8096,
   "unit": "score_0_5",
   "ci_low": 2.3877,
   "ci_high": 3.1941,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score_all:grok-4.7",
   "label": "grok-4.7, weighted score, all 10 tasks",
   "value": 2.8836,
   "unit": "score_0_5",
   "ci_low": 2.5762,
   "ci_high": 3.1668,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "cost:grok-4.7",
   "label": "grok-4.7, generation cost per task",
   "value": 0.0969,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:grok-4.7",
   "label": "grok-4.7, latency median",
   "value": 213.2583,
   "unit": "seconds",
   "ci_low": 172.2514,
   "ci_high": 238.9544,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:grok-4.7",
   "label": "grok-4.7, hard fails",
   "value": 1.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:grok-4.7",
   "label": "grok-4.7, accuracy",
   "value": 2.8333,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:grok-4.7",
   "label": "grok-4.7, grounding",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:grok-4.7",
   "label": "grok-4.7, specificity",
   "value": 3.3333,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:grok-4.7",
   "label": "grok-4.7, actionability",
   "value": 3.5,
   "unit": "score_0_5",
   "ci_low": 3.1667,
   "ci_high": 3.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:grok-4.7",
   "label": "grok-4.7, polish",
   "value": 2.5,
   "unit": "score_0_5",
   "ci_low": 2.0,
   "ci_high": 3.1667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:grok-4.7",
   "label": "grok-4.7, end to end (x2)",
   "value": 2.8333,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.0,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:gemini-4-argon",
   "label": "gemini-4-argon, weighted score, 6 common tasks",
   "value": 2.5716,
   "unit": "score_0_5",
   "ci_low": 2.2359,
   "ci_high": 2.9834,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score_all:gemini-4-argon",
   "label": "gemini-4-argon, weighted score, all 10 tasks",
   "value": 2.6032,
   "unit": "score_0_5",
   "ci_low": 2.3594,
   "ci_high": 2.8846,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "cost:gemini-4-argon",
   "label": "gemini-4-argon, generation cost per task",
   "value": 0.058,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:gemini-4-argon",
   "label": "gemini-4-argon, latency median",
   "value": 65.9383,
   "unit": "seconds",
   "ci_low": 49.8025,
   "ci_high": 74.1104,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:gemini-4-argon",
   "label": "gemini-4-argon, hard fails",
   "value": 1.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:gemini-4-argon",
   "label": "gemini-4-argon, accuracy",
   "value": 2.5,
   "unit": "score_0_5",
   "ci_low": 2.1667,
   "ci_high": 2.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:gemini-4-argon",
   "label": "gemini-4-argon, grounding",
   "value": 2.5,
   "unit": "score_0_5",
   "ci_low": 2.1667,
   "ci_high": 2.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:gemini-4-argon",
   "label": "gemini-4-argon, specificity",
   "value": 3.1667,
   "unit": "score_0_5",
   "ci_low": 2.6667,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:gemini-4-argon",
   "label": "gemini-4-argon, actionability",
   "value": 3.5,
   "unit": "score_0_5",
   "ci_low": 3.1667,
   "ci_high": 3.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:gemini-4-argon",
   "label": "gemini-4-argon, polish",
   "value": 2.6667,
   "unit": "score_0_5",
   "ci_low": 2.1667,
   "ci_high": 3.3333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:gemini-4-argon",
   "label": "gemini-4-argon, end to end (x2)",
   "value": 2.5,
   "unit": "score_0_5",
   "ci_low": 2.1667,
   "ci_high": 2.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:claude-haiku-5.5",
   "label": "claude-haiku-5.5, weighted score, 6 common tasks",
   "value": 2.2699,
   "unit": "score_0_5",
   "ci_low": 2.0195,
   "ci_high": 2.6786,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score_all:claude-haiku-5.5",
   "label": "claude-haiku-5.5, weighted score, all 10 tasks",
   "value": 2.5714,
   "unit": "score_0_5",
   "ci_low": 2.2366,
   "ci_high": 3.0,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "cost:claude-haiku-5.5",
   "label": "claude-haiku-5.5, generation cost per task",
   "value": 0.0048,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "latency_p50:claude-haiku-5.5",
   "label": "claude-haiku-5.5, latency median",
   "value": 35.2789,
   "unit": "seconds",
   "ci_low": 31.7778,
   "ci_high": 39.2224,
   "basis": "measured",
   "n": 6,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": "interval is the IQR, not a CI"
  },
  {
   "id": "hard_fails:claude-haiku-5.5",
   "label": "claude-haiku-5.5, hard fails",
   "value": 3.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:accuracy:claude-haiku-5.5",
   "label": "claude-haiku-5.5, accuracy",
   "value": 2.3333,
   "unit": "score_0_5",
   "ci_low": 2.0,
   "ci_high": 2.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:grounding:claude-haiku-5.5",
   "label": "claude-haiku-5.5, grounding",
   "value": 2.3333,
   "unit": "score_0_5",
   "ci_low": 1.6667,
   "ci_high": 2.8333,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:specificity:claude-haiku-5.5",
   "label": "claude-haiku-5.5, specificity",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.3333,
   "ci_high": 3.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:actionability:claude-haiku-5.5",
   "label": "claude-haiku-5.5, actionability",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:polish:claude-haiku-5.5",
   "label": "claude-haiku-5.5, polish",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.5,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "dim:e2e_wow:claude-haiku-5.5",
   "label": "claude-haiku-5.5, end to end (x2)",
   "value": 2.3333,
   "unit": "score_0_5",
   "ci_low": 2.0,
   "ci_high": 2.6667,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "score6:astra_production_prompt",
   "label": "gpt-6-astra on production QUALITY_BAR, 6 tasks",
   "value": 2.3648,
   "unit": "score_0_5",
   "ci_low": 1.918,
   "ci_high": 2.7946,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "heldout:production",
   "label": "held-out weighted score, production",
   "value": 2.7276,
   "unit": "score_0_5",
   "ci_low": 2.3775,
   "ci_high": 3.1468,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/optimizer/runs.csv; evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "heldout:v1",
   "label": "held-out weighted score, v1",
   "value": 3.5099,
   "unit": "score_0_5",
   "ci_low": 3.0156,
   "ci_high": 4.028,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/optimizer/runs.csv; evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "heldout:v2",
   "label": "held-out weighted score, v2",
   "value": 3.4436,
   "unit": "score_0_5",
   "ci_low": 3.0942,
   "ci_high": 3.8072,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/optimizer/runs.csv; evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "heldout:v3",
   "label": "held-out weighted score, v3",
   "value": 3.5616,
   "unit": "score_0_5",
   "ci_low": 3.3082,
   "ci_high": 3.8615,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/optimizer/runs.csv; evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "heldout:v4",
   "label": "held-out weighted score, v4",
   "value": 3.6447,
   "unit": "score_0_5",
   "ci_low": 3.286,
   "ci_high": 4.0682,
   "basis": "measured",
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/optimizer/runs.csv; evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:gpt-6-astra",
   "label": "gpt-6-astra, Advice and research",
   "value": 3.571,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:pascal-1-fallback",
   "label": "pascal-1-fallback, Advice and research",
   "value": 3.4284,
   "unit": "score_0_5",
   "ci_low": 2.857,
   "ci_high": 3.857,
   "basis": "measured",
   "n": 3,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:pascal-1",
   "label": "pascal-1, Advice and research",
   "value": 4.6284,
   "unit": "score_0_5",
   "ci_low": 4.0,
   "ci_high": 4.857,
   "basis": "measured",
   "n": 3,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:gpt-6.1-terra",
   "label": "gpt-6.1-terra, Advice and research",
   "value": 3.8286,
   "unit": "score_0_5",
   "ci_low": 3.429,
   "ci_high": 4.143,
   "basis": "measured",
   "n": 3,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:grok-4.7",
   "label": "grok-4.7, Advice and research",
   "value": 3.286,
   "unit": "score_0_5",
   "ci_low": 3.143,
   "ci_high": 3.429,
   "basis": "measured",
   "n": 3,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:gemini-4-argon",
   "label": "gemini-4-argon, Advice and research",
   "value": 2.9146,
   "unit": "score_0_5",
   "ci_low": 2.429,
   "ci_high": 3.429,
   "basis": "measured",
   "n": 3,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Advice and research:claude-haiku-5.5",
   "label": "claude-haiku-5.5, Advice and research",
   "value": 3.4286,
   "unit": "score_0_5",
   "ci_low": 3.143,
   "ci_high": 3.714,
   "basis": "measured",
   "n": 3,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Agent setup and behaviour:pascal-1-fallback",
   "label": "pascal-1-fallback, Agent setup and behaviour",
   "value": 3.143,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Agent setup and behaviour:pascal-1",
   "label": "pascal-1, Agent setup and behaviour",
   "value": 3.429,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Agent setup and behaviour:gpt-6.1-terra",
   "label": "gpt-6.1-terra, Agent setup and behaviour",
   "value": 3.286,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Agent setup and behaviour:grok-4.7",
   "label": "grok-4.7, Agent setup and behaviour",
   "value": 2.714,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Agent setup and behaviour:gemini-4-argon",
   "label": "gemini-4-argon, Agent setup and behaviour",
   "value": 2.571,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Agent setup and behaviour:claude-haiku-5.5",
   "label": "claude-haiku-5.5, Agent setup and behaviour",
   "value": 3.143,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:gpt-6-astra",
   "label": "gpt-6-astra, Drafting and redlining",
   "value": 3.3363,
   "unit": "score_0_5",
   "ci_low": 3.0589,
   "ci_high": 3.6519,
   "basis": "measured",
   "n": 5,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:pascal-1-fallback",
   "label": "pascal-1-fallback, Drafting and redlining",
   "value": 3.1581,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.2709,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:pascal-1",
   "label": "pascal-1, Drafting and redlining",
   "value": 2.9849,
   "unit": "score_0_5",
   "ci_low": 2.4543,
   "ci_high": 3.625,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:gpt-6.1-terra",
   "label": "gpt-6.1-terra, Drafting and redlining",
   "value": 2.8497,
   "unit": "score_0_5",
   "ci_low": 2.6056,
   "ci_high": 3.1698,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:grok-4.7",
   "label": "grok-4.7, Drafting and redlining",
   "value": 2.8045,
   "unit": "score_0_5",
   "ci_low": 2.4013,
   "ci_high": 3.1668,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:gemini-4-argon",
   "label": "gemini-4-argon, Drafting and redlining",
   "value": 2.5264,
   "unit": "score_0_5",
   "ci_low": 2.2182,
   "ci_high": 2.8949,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:ucg:Drafting and redlining:claude-haiku-5.5",
   "label": "claude-haiku-5.5, Drafting and redlining",
   "value": 2.2556,
   "unit": "score_0_5",
   "ci_low": 2.0195,
   "ci_high": 2.6223,
   "basis": "measured",
   "n": 6,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:CA:pascal-1-fallback",
   "label": "pascal-1-fallback, CA",
   "value": 3.857,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:CA:pascal-1",
   "label": "pascal-1, CA",
   "value": 4.714,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:CA:gpt-6.1-terra",
   "label": "gpt-6.1-terra, CA",
   "value": 3.714,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:CA:grok-4.7",
   "label": "grok-4.7, CA",
   "value": 3.429,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:CA:gemini-4-argon",
   "label": "gemini-4-argon, CA",
   "value": 3.143,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:CA:claude-haiku-5.5",
   "label": "claude-haiku-5.5, CA",
   "value": 3.714,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:gpt-6-astra",
   "label": "gpt-6-astra, DE",
   "value": 3.439,
   "unit": "score_0_5",
   "ci_low": 3.2043,
   "ci_high": 3.8215,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:pascal-1-fallback",
   "label": "pascal-1-fallback, DE",
   "value": 3.1328,
   "unit": "score_0_5",
   "ci_low": 2.9428,
   "ci_high": 3.286,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:pascal-1",
   "label": "pascal-1, DE",
   "value": 2.9286,
   "unit": "score_0_5",
   "ci_low": 2.2855,
   "ci_high": 3.786,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:gpt-6.1-terra",
   "label": "gpt-6.1-terra, DE",
   "value": 2.8062,
   "unit": "score_0_5",
   "ci_low": 2.5,
   "ci_high": 3.2145,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:grok-4.7",
   "label": "grok-4.7, DE",
   "value": 2.7347,
   "unit": "score_0_5",
   "ci_low": 2.2142,
   "ci_high": 3.253,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:gemini-4-argon",
   "label": "gemini-4-argon, DE",
   "value": 2.5715,
   "unit": "score_0_5",
   "ci_low": 2.1785,
   "ci_high": 3.0718,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:DE:claude-haiku-5.5",
   "label": "claude-haiku-5.5, DE",
   "value": 2.2143,
   "unit": "score_0_5",
   "ci_low": 2.0,
   "ci_high": 2.6923,
   "basis": "measured",
   "n": 4,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:EU:pascal-1-fallback",
   "label": "pascal-1-fallback, EU",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": 2.857,
   "ci_high": 3.143,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:EU:pascal-1",
   "label": "pascal-1, EU",
   "value": 4.214,
   "unit": "score_0_5",
   "ci_low": 3.571,
   "ci_high": 4.857,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:EU:gpt-6.1-terra",
   "label": "gpt-6.1-terra, EU",
   "value": 3.643,
   "unit": "score_0_5",
   "ci_low": 3.143,
   "ci_high": 4.143,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:EU:grok-4.7",
   "label": "grok-4.7, EU",
   "value": 3.0715,
   "unit": "score_0_5",
   "ci_low": 3.0,
   "ci_high": 3.143,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:EU:gemini-4-argon",
   "label": "gemini-4-argon, EU",
   "value": 2.5,
   "unit": "score_0_5",
   "ci_low": 2.429,
   "ci_high": 2.571,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:EU:claude-haiku-5.5",
   "label": "claude-haiku-5.5, EU",
   "value": 2.9285,
   "unit": "score_0_5",
   "ci_low": 2.714,
   "ci_high": 3.143,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:gpt-6-astra",
   "label": "gpt-6-astra, Multi",
   "value": 3.571,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:pascal-1-fallback",
   "label": "pascal-1-fallback, Multi",
   "value": 3.2857,
   "unit": "score_0_5",
   "ci_low": 3.143,
   "ci_high": 3.714,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:pascal-1",
   "label": "pascal-1, Multi",
   "value": 3.5717,
   "unit": "score_0_5",
   "ci_low": 3.429,
   "ci_high": 4.0,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:gpt-6.1-terra",
   "label": "gpt-6.1-terra, Multi",
   "value": 3.3218,
   "unit": "score_0_5",
   "ci_low": 3.286,
   "ci_high": 3.429,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:grok-4.7",
   "label": "grok-4.7, Multi",
   "value": 2.857,
   "unit": "score_0_5",
   "ci_low": 2.714,
   "ci_high": 3.286,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:gemini-4-argon",
   "label": "gemini-4-argon, Multi",
   "value": 2.7855,
   "unit": "score_0_5",
   "ci_low": 2.571,
   "ci_high": 3.429,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:Multi:claude-haiku-5.5",
   "label": "claude-haiku-5.5, Multi",
   "value": 3.2145,
   "unit": "score_0_5",
   "ci_low": 3.143,
   "ci_high": 3.429,
   "basis": "measured",
   "n": 2,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:gpt-6-astra",
   "label": "gpt-6-astra, NY",
   "value": 2.857,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:pascal-1-fallback",
   "label": "pascal-1-fallback, NY",
   "value": 3.286,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:pascal-1",
   "label": "pascal-1, NY",
   "value": 2.857,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:gpt-6.1-terra",
   "label": "gpt-6.1-terra, NY",
   "value": 2.857,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:grok-4.7",
   "label": "grok-4.7, NY",
   "value": 3.0,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:gemini-4-argon",
   "label": "gemini-4-argon, NY",
   "value": 2.286,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "split:jur:NY:claude-haiku-5.5",
   "label": "claude-haiku-5.5, NY",
   "value": 2.143,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 1,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/results.csv",
   "note": null
  },
  {
   "id": "citations:exists",
   "label": "citations that exist",
   "value": 0.9931,
   "unit": "share",
   "ci_low": 0.9854,
   "ci_high": 1.0,
   "basis": "measured",
   "n": 433,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": "'unsure' left out of the denominator"
  },
  {
   "id": "citations:provision",
   "label": "citations on the right provision",
   "value": 0.986,
   "unit": "share",
   "ci_low": 0.9736,
   "ci_high": 0.9955,
   "basis": "measured",
   "n": 429,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": "'unsure' left out of the denominator"
  },
  {
   "id": "citations:full",
   "label": "citations fully accurate",
   "value": 0.7179,
   "unit": "share",
   "ci_low": 0.6667,
   "ci_high": 0.7695,
   "basis": "measured",
   "n": 429,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": "'unsure' left out of the denominator"
  },
  {
   "id": "citations:partly",
   "label": "citations partly accurate",
   "value": 0.2681,
   "unit": "share",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 429,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:wrong",
   "label": "citations wrong",
   "value": 0.014,
   "unit": "share",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 429,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:fabricated",
   "label": "citations that do not exist",
   "value": 3.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 438,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:grok-4.7",
   "label": "grok-4.7, citations fully accurate",
   "value": 0.6622,
   "unit": "share",
   "ci_low": 0.5517,
   "ci_high": 0.8033,
   "basis": "measured",
   "n": 74,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:pascal-1",
   "label": "pascal-1, citations fully accurate",
   "value": 0.7971,
   "unit": "share",
   "ci_low": 0.7123,
   "ci_high": 0.9107,
   "basis": "measured",
   "n": 69,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:gpt-6-astra",
   "label": "gpt-6-astra, citations fully accurate",
   "value": 0.8421,
   "unit": "share",
   "ci_low": 0.7059,
   "ci_high": 1.0,
   "basis": "measured",
   "n": 19,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:gemini-4-argon",
   "label": "gemini-4-argon, citations fully accurate",
   "value": 0.7105,
   "unit": "share",
   "ci_low": 0.5312,
   "ci_high": 0.8571,
   "basis": "measured",
   "n": 76,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:pascal-1-fallback",
   "label": "pascal-1-fallback, citations fully accurate",
   "value": 0.6818,
   "unit": "share",
   "ci_low": 0.5641,
   "ci_high": 0.7767,
   "basis": "measured",
   "n": 88,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:gpt-6.1-terra",
   "label": "gpt-6.1-terra, citations fully accurate",
   "value": 0.7778,
   "unit": "share",
   "ci_low": 0.6786,
   "ci_high": 0.8537,
   "basis": "measured",
   "n": 45,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "citations:full:claude-haiku-5.5",
   "label": "claude-haiku-5.5, citations fully accurate",
   "value": 0.6724,
   "unit": "share",
   "ci_low": 0.4783,
   "ci_high": 0.8103,
   "basis": "measured",
   "n": 58,
   "judge": "openai/gpt-6.1-terra (fetched) and pascal-1 (model-checked)",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/citations.csv",
   "note": null
  },
  {
   "id": "judge:pearson",
   "label": "primary vs second judge, Pearson r",
   "value": 0.3085,
   "unit": "r",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 42,
   "judge": "pascal-1 vs gemini-4-argon",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/xjudge.csv",
   "note": null
  },
  {
   "id": "judge:spearman",
   "label": "primary vs second judge, Spearman rho",
   "value": 0.2853,
   "unit": "rho",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 42,
   "judge": "pascal-1 vs gemini-4-argon",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/xjudge.csv",
   "note": null
  },
  {
   "id": "judge:mad",
   "label": "primary vs second judge, mean absolute difference",
   "value": 1.7686,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 42,
   "judge": "pascal-1 vs gemini-4-argon",
   "date": "2026-10-31",
   "source": "evals/runs/2026-10-31-measured/xjudge.csv",
   "note": null
  },
  {
   "id": "judge:retest_mad",
   "label": "primary judge test-retest, mean absolute difference",
   "value": 0.06,
   "unit": "score_0_5",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 7,
   "judge": "pascal-1",
   "date": "2026-10-31",
   "source": "evals/model-bakeoff.md",
   "note": "reported; raw retest rows not on disk, not recomputed"
  },
  {
   "id": "judge:human_agreement",
   "label": "judge agreement with lawyer labels",
   "value": null,
   "unit": "kappa",
   "ci_low": null,
   "ci_high": null,
   "basis": "not_run",
   "n": null,
   "judge": null,
   "date": "2026-10-31",
   "source": "",
   "note": "not yet measured"
  },
  {
   "id": "routing:regex:agent_ok",
   "label": "regex router, agent_ok",
   "value": 0.4639,
   "unit": "share",
   "ci_low": 0.3679,
   "ci_high": 0.5626,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/routing/runs/route-only-v2-97.json",
   "note": null
  },
  {
   "id": "routing:regex:instrument_ok",
   "label": "regex router, instrument_ok",
   "value": 0.7216,
   "unit": "share",
   "ci_low": 0.6253,
   "ci_high": 0.8011,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/routing/runs/route-only-v2-97.json",
   "note": null
  },
  {
   "id": "routing:regex:amounts_ok",
   "label": "regex router, amounts_ok",
   "value": 0.6495,
   "unit": "share",
   "ci_low": 0.5505,
   "ci_high": 0.7371,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/routing/runs/route-only-v2-97.json",
   "note": null
  },
  {
   "id": "routing:regex:route_pass",
   "label": "regex router, route_pass",
   "value": 0.2268,
   "unit": "share",
   "ci_low": 0.1548,
   "ci_high": 0.3196,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/routing/runs/route-only-v2-97.json",
   "note": null
  },
  {
   "id": "routing:llm:task",
   "label": "LLM intent step, task",
   "value": 0.9381,
   "unit": "share",
   "ci_low": 0.8716,
   "ci_high": 0.9713,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "note": null
  },
  {
   "id": "routing:llm:doc_type",
   "label": "LLM intent step, doc_type",
   "value": 0.9485,
   "unit": "share",
   "ci_low": 0.885,
   "ci_high": 0.9778,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "note": null
  },
  {
   "id": "routing:llm:no_dur_money",
   "label": "LLM intent step, no_dur_money",
   "value": 1.0,
   "unit": "share",
   "ci_low": 0.9619,
   "ci_high": 1.0,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "note": null
  },
  {
   "id": "routing:llm:pass",
   "label": "LLM intent step, pass",
   "value": 0.4948,
   "unit": "share",
   "ci_low": 0.3975,
   "ci_high": 0.5926,
   "basis": "measured",
   "n": 97,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "note": null
  },
  {
   "id": "drafting:main_pascal-1",
   "label": "drafting pass, production prompt, pascal-1",
   "value": 0.0,
   "unit": "share",
   "ci_low": 0.0,
   "ci_high": 0.1246,
   "basis": "measured",
   "n": 27,
   "judge": "deterministic + Pascal-1",
   "date": "2026-10-31",
   "source": "evals/routing/runs/",
   "note": null
  },
  {
   "id": "drafting:main_astra",
   "label": "drafting pass, production prompt, gpt-6-astra",
   "value": 0.0,
   "unit": "share",
   "ci_low": 0.0,
   "ci_high": 0.4899,
   "basis": "measured",
   "n": 4,
   "judge": "deterministic + Pascal-1",
   "date": "2026-10-31",
   "source": "evals/routing/runs/",
   "note": null
  },
  {
   "id": "drafting:eng_pascal-1",
   "label": "drafting pass, request passed through, pascal-1",
   "value": 0.7,
   "unit": "share",
   "ci_low": 0.3968,
   "ci_high": 0.8922,
   "basis": "measured",
   "n": 10,
   "judge": "deterministic + Pascal-1",
   "date": "2026-10-31",
   "source": "evals/routing/runs/",
   "note": null
  },
  {
   "id": "drafting:fix_pascal-1",
   "label": "drafting pass, proposed fix v1, pascal-1",
   "value": 0.963,
   "unit": "share",
   "ci_low": 0.8172,
   "ci_high": 0.9934,
   "basis": "measured",
   "n": 27,
   "judge": "deterministic + Pascal-1",
   "date": "2026-10-31",
   "source": "evals/routing/runs/",
   "note": null
  },
  {
   "id": "drafting:fix_v2_pascal-1",
   "label": "drafting pass, proposed fix v2, pascal-1",
   "value": 1.0,
   "unit": "share",
   "ci_low": 0.5101,
   "ci_high": 1.0,
   "basis": "measured",
   "n": 4,
   "judge": "deterministic + Pascal-1",
   "date": "2026-10-31",
   "source": "evals/routing/runs/",
   "note": null
  },
  {
   "id": "lint:failures_before",
   "label": "prompt lint failures before hardening",
   "value": 71.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 55,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/proposed/hardening/lint-before.json",
   "note": null
  },
  {
   "id": "lint:failures_after",
   "label": "prompt lint failures after hardening (proposal, not applied)",
   "value": 0.0,
   "unit": "count",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 55,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/proposed/hardening/lint-after.json",
   "note": null
  },
  {
   "id": "selfserve:measured",
   "label": "self-serve rate, measured",
   "value": 0.0612,
   "unit": "share",
   "ci_low": 0.021,
   "ci_high": 0.1652,
   "basis": "measured",
   "n": 49,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/selfserve/self-serve.json",
   "note": null
  },
  {
   "id": "selfserve:estimated",
   "label": "self-serve rate, estimated after queued fixes, not re-run",
   "value": 0.7931,
   "unit": "share",
   "ci_low": null,
   "ci_high": null,
   "basis": "estimated",
   "n": 58,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/selfserve/self-serve.json",
   "note": "upper bound"
  },
  {
   "id": "selfserve:cost_avoided_measured",
   "label": "lawyer cost avoided, measured",
   "value": 385.72,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "measured",
   "n": 49,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/selfserve/self-serve.json",
   "note": null
  },
  {
   "id": "selfserve:cost_avoided_estimated",
   "label": "lawyer cost avoided, estimated, not re-run",
   "value": 13952.7,
   "unit": "usd",
   "ci_low": null,
   "ci_high": null,
   "basis": "estimated",
   "n": 58,
   "judge": null,
   "date": "2026-10-31",
   "source": "evals/selfserve/self-serve.json",
   "note": "upper bound"
  }
 ],
 "significance": [
  {
   "id": "astra_vs_claude-haiku-5.5",
   "a": "openai/gpt-6-astra",
   "b": "anthropic/claude-haiku-5.5",
   "delta": 1.0794,
   "ci_low": 0.7855,
   "ci_high": 1.286,
   "p_bootstrap": 0.0,
   "p_permutation": 0.0312,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": 0.1875
  },
  {
   "id": "astra_vs_pascal-1-fallback",
   "a": "openai/gpt-6-astra",
   "b": "pascal-1-fallback",
   "delta": 0.1587,
   "ci_low": -0.143,
   "ci_high": 0.4285,
   "p_bootstrap": 0.3086,
   "p_permutation": 0.375,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": 0.75
  },
  {
   "id": "astra_vs_gemini-4-argon",
   "a": "openai/gpt-6-astra",
   "b": "google/gemini-4-argon",
   "delta": 0.7778,
   "ci_low": 0.5808,
   "ci_high": 0.9591,
   "p_bootstrap": 0.0,
   "p_permutation": 0.0312,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": 0.1875
  },
  {
   "id": "astra_vs_pascal-1",
   "a": "openai/gpt-6-astra",
   "b": "pascal-1",
   "delta": 0.3732,
   "ci_low": -0.1532,
   "ci_high": 0.8266,
   "p_bootstrap": 0.168,
   "p_permutation": 0.5,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": 0.75
  },
  {
   "id": "astra_vs_gpt-6.1-terra",
   "a": "openai/gpt-6-astra",
   "b": "openai/gpt-6.1-terra",
   "delta": 0.5001,
   "ci_low": 0.2039,
   "ci_high": 0.7663,
   "p_bootstrap": 0.0002,
   "p_permutation": 0.0625,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": 0.25
  },
  {
   "id": "astra_vs_grok-4.7",
   "a": "openai/gpt-6-astra",
   "b": "x-ai/grok-4.7",
   "delta": 0.5398,
   "ci_low": 0.1846,
   "ci_high": 0.8788,
   "p_bootstrap": 0.0034,
   "p_permutation": 0.0938,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": 0.2812
  },
  {
   "id": "astra_v26_vs_astra_production",
   "a": "gpt-6-astra + pack v2.6",
   "b": "gpt-6-astra + production QUALITY_BAR",
   "delta": 0.9845,
   "ci_low": 0.597,
   "ci_high": 1.3433,
   "p_bootstrap": 0.0,
   "p_permutation": 0.0312,
   "n_tasks": 6,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": null
  },
  {
   "id": "heldout_v4_vs_production",
   "a": "pascal-1 + optimized v4",
   "b": "pascal-1 + production QUALITY_BAR",
   "delta": 0.9171,
   "ci_low": 0.6753,
   "ci_high": 1.1817,
   "p_bootstrap": 0.0,
   "p_permutation": 0.002,
   "n_tasks": 10,
   "test": "paired percentile bootstrap over tasks (two-sided) and exact sign-flip permutation, task-weighted",
   "p_holm": null
  }
 ],
 "charts": [
  {
   "id": "scorecard",
   "figure": 0,
   "title": "Scorecard",
   "subtitle": "A legal work benchmark by Index Law",
   "headline": {
    "value": 3.35,
    "label": "Pascal on gpt-6-astra",
    "n": 6
   },
   "kind": "scorecard",
   "png": "charts/00-scorecard.png",
   "svg": "charts/00-scorecard.svg",
   "series": [
    {
     "id": "cards",
     "label": "cards",
     "points": [
      {
       "key": "card0",
       "label": "pascal on gpt-6-astra",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "display": "3.35",
        "detail": "of 5, 95% CI 3.09 to 3.66; 6 tasks, judge pascal-1"
       }
      },
      {
       "key": "card1",
       "label": "citations fully accurate",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "display": "71.8%",
        "detail": "438 checked; 3 do not exist"
       }
      },
      {
       "key": "card2",
       "label": "held-out, prompt optimized",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "display": "2.73 to 3.64",
        "detail": "10 tasks, permutation p=0.002; selection on held-out"
       }
      },
      {
       "key": "card3",
       "label": "llm intent, task right",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "display": "91/97",
        "detail": "regex router 45/97; same cases"
       }
      },
      {
       "key": "card4",
       "label": "self-serve, measured",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "display": "6.1%",
        "detail": "3 of 49 tasks; estimates not shown"
       }
      }
     ]
    }
   ],
   "footnote": "",
   "status": "ready",
   "source": [],
   "n": null,
   "judge": "pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "headline",
   "figure": 1,
   "title": "Overall quality by model",
   "subtitle": "Pascal v2.6 pack, 6 common tasks, task weighted",
   "headline": {
    "value": 3.35,
    "label": "Pascal on gpt-6-astra",
    "n": 6
   },
   "kind": "bar_ci",
   "png": "charts/01-headline-quality.png",
   "svg": "charts/01-headline-quality.svg",
   "series": [
    {
     "id": "models",
     "label": "Pascal v2.6 pack",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 3.3493,
       "ci_low": 3.0873,
       "ci_high": 3.6609,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": true
       }
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.1906,
       "ci_low": 3.0204,
       "ci_high": 3.3518,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": false
       }
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 2.9762,
       "ci_low": 2.4215,
       "ci_high": 3.694,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": false
       }
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 2.8493,
       "ci_low": 2.593,
       "ci_high": 3.2002,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": false
       }
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 2.8096,
       "ci_low": 2.3877,
       "ci_high": 3.1941,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": false
       }
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 2.5716,
       "ci_low": 2.2359,
       "ci_high": 2.9834,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": false
       }
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 2.2699,
       "ci_low": 2.0195,
       "ci_high": 2.6786,
       "n": 6,
       "basis": "measured",
       "extra": {
        "highlight": false
       }
      }
     ]
    },
    {
     "id": "production",
     "label": "production prompt (QUALITY_BAR)",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra, production prompt",
       "x": null,
       "y": 2.3648,
       "ci_low": 1.918,
       "ci_high": 2.7946,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Judge pascal-1 is also a contestant. 95% CI by bootstrap over tasks. Paired exact permutation vs astra: claude-haiku-5.5 p=0.031, gemini-4-argon p=0.031, gpt-6.1-terra p=0.062, grok-4.7 p=0.094, pascal-1-fallback p=0.375, pascal-1 p=0.500; none survive Holm correction for 6 comparisons.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/results.csv"
   ],
   "n": 42,
   "judge": "pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "dimensions",
   "figure": 2,
   "title": "Per-dimension scores",
   "subtitle": "6 common tasks",
   "headline": {
    "value": 2.9,
    "label": "mean accuracy, all models",
    "n": 42
   },
   "kind": "small_multiples",
   "png": "charts/02-dimensions-small-multiples.png",
   "svg": "charts/02-dimensions-small-multiples.svg",
   "series": [
    {
     "id": "accuracy",
     "label": "Accuracy",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 3.3333,
       "ci_low": 3.0,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.3333,
       "ci_low": 3.0,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 3.0,
       "ci_low": 3.0,
       "ci_high": 3.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 2.8333,
       "ci_low": 2.5,
       "ci_high": 3.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 2.5,
       "ci_low": 2.1667,
       "ci_high": 2.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 2.3333,
       "ci_low": 2.0,
       "ci_high": 2.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "grounding",
     "label": "Grounding",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 4.0,
       "ci_low": 4.0,
       "ci_high": 4.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.0,
       "ci_low": 3.0,
       "ci_high": 3.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 3.6667,
       "ci_low": 2.8333,
       "ci_high": 4.3333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 3.3333,
       "ci_low": 2.6667,
       "ci_high": 3.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 2.5,
       "ci_low": 2.1667,
       "ci_high": 2.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 2.3333,
       "ci_low": 1.6667,
       "ci_high": 2.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "specificity",
     "label": "Specificity",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 3.6667,
       "ci_low": 3.1667,
       "ci_high": 4.3333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.5,
       "ci_low": 3.1667,
       "ci_high": 3.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 3.5,
       "ci_low": 2.6667,
       "ci_high": 4.3333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 3.1667,
       "ci_low": 2.6667,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 3.3333,
       "ci_low": 3.0,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 3.1667,
       "ci_low": 2.6667,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 3.0,
       "ci_low": 2.3333,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "actionability",
     "label": "Actionability",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 3.6667,
       "ci_low": 3.3333,
       "ci_high": 4.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.6667,
       "ci_low": 3.3333,
       "ci_high": 4.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 3.6667,
       "ci_low": 3.1667,
       "ci_high": 4.3333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 3.1667,
       "ci_low": 2.6667,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 3.5,
       "ci_low": 3.1667,
       "ci_high": 3.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 3.5,
       "ci_low": 3.1667,
       "ci_high": 3.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "polish",
     "label": "Polish",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.3333,
       "ci_low": 3.0,
       "ci_high": 3.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 3.1667,
       "ci_low": 3.0,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 2.5,
       "ci_low": 2.0,
       "ci_high": 3.1667,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 2.6667,
       "ci_low": 2.1667,
       "ci_high": 3.3333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "e2e_wow",
     "label": "End to end (x2)",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 3.0,
       "ci_low": 2.5,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 2.8333,
       "ci_low": 2.1667,
       "ci_high": 3.5,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 2.5,
       "ci_low": 2.1667,
       "ci_high": 2.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 2.8333,
       "ci_low": 2.5,
       "ci_high": 3.0,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 2.5,
       "ci_low": 2.1667,
       "ci_high": 2.8333,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 2.3333,
       "ci_low": 2.0,
       "ci_high": 2.6667,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Judge pascal-1 is also a contestant.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/results.csv"
   ],
   "n": 42,
   "judge": "pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "splits",
   "figure": 3,
   "title": "Scores by use case and jurisdiction",
   "subtitle": "all bake-off tasks",
   "headline": null,
   "kind": "dot_ci",
   "png": "charts/03-splits-usecase-jurisdiction.png",
   "svg": "charts/03-splits-usecase-jurisdiction.svg",
   "series": [
    {
     "id": "ucg:Advice and research",
     "label": "Advice and research",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": "Advice and research",
       "y": 3.571,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "Advice and research",
       "y": 3.4284,
       "ci_low": 2.857,
       "ci_high": 3.857,
       "n": 3,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "Advice and research",
       "y": 4.6284,
       "ci_low": 4.0,
       "ci_high": 4.857,
       "n": 3,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "Advice and research",
       "y": 3.8286,
       "ci_low": 3.429,
       "ci_high": 4.143,
       "n": 3,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "Advice and research",
       "y": 3.286,
       "ci_low": 3.143,
       "ci_high": 3.429,
       "n": 3,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "Advice and research",
       "y": 2.9146,
       "ci_low": 2.429,
       "ci_high": 3.429,
       "n": 3,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "Advice and research",
       "y": 3.4286,
       "ci_low": 3.143,
       "ci_high": 3.714,
       "n": 3,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "ucg:Agent setup and behaviour",
     "label": "Agent setup and behaviour",
     "points": [
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "Agent setup and behaviour",
       "y": 3.143,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "Agent setup and behaviour",
       "y": 3.429,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "Agent setup and behaviour",
       "y": 3.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "Agent setup and behaviour",
       "y": 2.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "Agent setup and behaviour",
       "y": 2.571,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "Agent setup and behaviour",
       "y": 3.143,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "ucg:Drafting and redlining",
     "label": "Drafting and redlining",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": "Drafting and redlining",
       "y": 3.3363,
       "ci_low": 3.0589,
       "ci_high": 3.664,
       "n": 5,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "Drafting and redlining",
       "y": 3.1581,
       "ci_low": 3.0068,
       "ci_high": 3.2709,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "Drafting and redlining",
       "y": 2.9849,
       "ci_low": 2.449,
       "ci_high": 3.625,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "Drafting and redlining",
       "y": 2.8497,
       "ci_low": 2.5974,
       "ci_high": 3.1682,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "Drafting and redlining",
       "y": 2.8045,
       "ci_low": 2.4013,
       "ci_high": 3.1589,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "Drafting and redlining",
       "y": 2.5264,
       "ci_low": 2.2145,
       "ci_high": 2.8774,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "Drafting and redlining",
       "y": 2.2556,
       "ci_low": 2.0195,
       "ci_high": 2.6134,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "jur:CA",
     "label": "CA",
     "points": [
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "CA",
       "y": 3.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "CA",
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "CA",
       "y": 3.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "CA",
       "y": 3.429,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "CA",
       "y": 3.143,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "CA",
       "y": 3.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "jur:DE",
     "label": "DE",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": "DE",
       "y": 3.439,
       "ci_low": 3.2043,
       "ci_high": 3.8215,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "DE",
       "y": 3.1328,
       "ci_low": 2.9428,
       "ci_high": 3.286,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "DE",
       "y": 2.9286,
       "ci_low": 2.2855,
       "ci_high": 3.786,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "DE",
       "y": 2.8062,
       "ci_low": 2.5,
       "ci_high": 3.2145,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "DE",
       "y": 2.7347,
       "ci_low": 2.2142,
       "ci_high": 3.253,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "DE",
       "y": 2.5715,
       "ci_low": 2.1785,
       "ci_high": 3.0718,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "DE",
       "y": 2.2143,
       "ci_low": 2.0,
       "ci_high": 2.6923,
       "n": 4,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "jur:EU",
     "label": "EU",
     "points": [
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "EU",
       "y": 3.0,
       "ci_low": 2.857,
       "ci_high": 3.143,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "EU",
       "y": 4.214,
       "ci_low": 3.571,
       "ci_high": 4.857,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "EU",
       "y": 3.643,
       "ci_low": 3.143,
       "ci_high": 4.143,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "EU",
       "y": 3.0715,
       "ci_low": 3.0,
       "ci_high": 3.143,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "EU",
       "y": 2.5,
       "ci_low": 2.429,
       "ci_high": 2.571,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "EU",
       "y": 2.9285,
       "ci_low": 2.714,
       "ci_high": 3.143,
       "n": 2,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "jur:Multi",
     "label": "Multi",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": "Multi",
       "y": 3.571,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "Multi",
       "y": 3.2857,
       "ci_low": 3.143,
       "ci_high": 3.714,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "Multi",
       "y": 3.5717,
       "ci_low": 3.429,
       "ci_high": 4.0,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "Multi",
       "y": 3.3218,
       "ci_low": 3.286,
       "ci_high": 3.429,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "Multi",
       "y": 2.857,
       "ci_low": 2.714,
       "ci_high": 3.286,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "Multi",
       "y": 2.7855,
       "ci_low": 2.571,
       "ci_high": 3.429,
       "n": 2,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "Multi",
       "y": 3.2145,
       "ci_low": 3.143,
       "ci_high": 3.429,
       "n": 2,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "jur:NY",
     "label": "NY",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": "NY",
       "y": 2.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": "NY",
       "y": 3.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": "NY",
       "y": 2.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": "NY",
       "y": 2.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": "NY",
       "y": 3.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": "NY",
       "y": 2.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": "NY",
       "y": 2.143,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Small cells. Judge pascal-1 is also a contestant.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/results.csv",
    "evals/matrix/coverage.csv"
   ],
   "n": 66,
   "judge": "pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "citations",
   "figure": 4,
   "title": "Citation accuracy by model",
   "subtitle": "fully accurate, partly accurate, wrong",
   "headline": {
    "value": 0.7179,
    "label": "fully accurate",
    "n": 438
   },
   "kind": "stacked_bar",
   "png": "charts/04-citation-accuracy.png",
   "svg": "charts/04-citation-accuracy.svg",
   "series": [
    {
     "id": "full",
     "label": "fully accurate",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 0.8421,
       "ci_low": 0.7059,
       "ci_high": 1.0,
       "n": 19,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 0.7971,
       "ci_low": 0.7123,
       "ci_high": 0.9107,
       "n": 69,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 0.7778,
       "ci_low": 0.6786,
       "ci_high": 0.8537,
       "n": 45,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 0.7105,
       "ci_low": 0.5312,
       "ci_high": 0.8571,
       "n": 76,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 0.6818,
       "ci_low": 0.5641,
       "ci_high": 0.7767,
       "n": 88,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 0.6724,
       "ci_low": 0.4783,
       "ci_high": 0.8103,
       "n": 58,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 0.6622,
       "ci_low": 0.5517,
       "ci_high": 0.8033,
       "n": 74,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "partly",
     "label": "partly accurate",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 0.1579,
       "ci_low": null,
       "ci_high": null,
       "n": 19,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 0.1884,
       "ci_low": null,
       "ci_high": null,
       "n": 69,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 0.2,
       "ci_low": null,
       "ci_high": null,
       "n": 45,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 0.2632,
       "ci_low": null,
       "ci_high": null,
       "n": 76,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 0.3068,
       "ci_low": null,
       "ci_high": null,
       "n": 88,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 0.3103,
       "ci_low": null,
       "ci_high": null,
       "n": 58,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 0.3378,
       "ci_low": null,
       "ci_high": null,
       "n": 74,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "wrong",
     "label": "wrong",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": 19,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 0.0145,
       "ci_low": null,
       "ci_high": null,
       "n": 69,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 0.0222,
       "ci_low": null,
       "ci_high": null,
       "n": 45,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 0.0263,
       "ci_low": null,
       "ci_high": null,
       "n": 76,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 0.0114,
       "ci_low": null,
       "ci_high": null,
       "n": 88,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 0.0172,
       "ci_low": null,
       "ci_high": null,
       "n": 58,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": 74,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "overall",
     "label": "all models",
     "points": [
      {
       "key": "exists",
       "label": "exist",
       "x": null,
       "y": 0.9931,
       "ci_low": 0.9854,
       "ci_high": 1.0,
       "n": 433,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "provision",
       "label": "right provision",
       "x": null,
       "y": 0.986,
       "ci_low": 0.9736,
       "ci_high": 0.9955,
       "n": 429,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "full",
       "label": "fully accurate",
       "x": null,
       "y": 0.7179,
       "ci_low": 0.6667,
       "ci_high": 0.7695,
       "n": 429,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "fabricated",
       "label": "do not exist (count)",
       "x": null,
       "y": 3.0,
       "ci_low": null,
       "ci_high": null,
       "n": 438,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Checker gpt-6.1-terra on fetched official text; pascal-1 for the rest.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/citations.csv"
   ],
   "n": 438,
   "judge": "gpt-6.1-terra; pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "cost_quality",
   "figure": 5,
   "title": "Quality vs cost per task",
   "subtitle": "6 common tasks",
   "headline": {
    "value": 0.248,
    "label": "USD per task, gpt-6-astra",
    "n": 6
   },
   "kind": "scatter_ci",
   "png": "charts/05-quality-vs-cost.png",
   "svg": "charts/05-quality-vs-cost.svg",
   "series": [
    {
     "id": "models",
     "label": "models",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": 0.24834,
       "y": 3.3493,
       "ci_low": 3.0873,
       "ci_high": 3.6609,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": 0.15913,
       "y": 3.1906,
       "ci_low": 3.0204,
       "ci_high": 3.3518,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": 0.04296,
       "y": 2.9762,
       "ci_low": 2.4215,
       "ci_high": 3.694,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": 0.01818,
       "y": 2.8493,
       "ci_low": 2.593,
       "ci_high": 3.2002,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": 0.09689,
       "y": 2.8096,
       "ci_low": 2.3877,
       "ci_high": 3.1941,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": 0.05803,
       "y": 2.5716,
       "ci_low": 2.2359,
       "ci_high": 2.9834,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": 0.0048,
       "y": 2.2699,
       "ci_low": 2.0195,
       "ci_high": 2.6786,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Cost from OpenRouter usage.cost. Judge pascal-1 is also a contestant.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/results.csv"
   ],
   "n": 42,
   "judge": "pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "cost_latency",
   "figure": 6,
   "title": "Cost and latency per task",
   "subtitle": "6 common tasks",
   "headline": {
    "value": 93.4,
    "label": "median seconds, gpt-6-astra",
    "n": 6
   },
   "kind": "bar",
   "png": "charts/06-cost-latency.png",
   "svg": "charts/06-cost-latency.svg",
   "series": [
    {
     "id": "cost",
     "label": "USD per task (mean)",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 0.2483,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 0.1591,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 0.043,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 0.0182,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 0.0969,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 0.058,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 0.0048,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "latency",
     "label": "seconds, median with IQR",
     "points": [
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 93.42,
       "ci_low": 89.4456,
       "ci_high": 97.181,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 61.9087,
       "ci_low": 55.4838,
       "ci_high": 68.1381,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 66.2052,
       "ci_low": 64.1051,
       "ci_high": 68.0281,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 37.4375,
       "ci_low": 29.9232,
       "ci_high": 39.9488,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 213.2583,
       "ci_low": 172.2514,
       "ci_high": 238.9544,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      },
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 65.9383,
       "ci_low": 49.8025,
       "ci_high": 74.1104,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 35.2789,
       "ci_low": 31.7778,
       "ci_high": 39.2224,
       "n": 6,
       "basis": "measured",
       "extra": {
        "interval": "IQR"
       }
      }
     ]
    }
   ],
   "footnote": "Generation cost and wall clock latency.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/results.csv"
   ],
   "n": 42,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "heldout",
   "figure": 7,
   "title": "Production prompt vs optimized prompt, held-out",
   "subtitle": "10 held-out tasks",
   "headline": {
    "value": 3.64,
    "label": "held-out score, v4",
    "n": 10
   },
   "kind": "bar_ci",
   "png": "charts/07-heldout-before-after.png",
   "svg": "charts/07-heldout-before-after.svg",
   "series": [
    {
     "id": "versions",
     "label": "prompt versions",
     "points": [
      {
       "key": "production",
       "label": "Production prompt (QUALITY_BAR)",
       "x": null,
       "y": 2.7276,
       "ci_low": 2.3775,
       "ci_high": 3.1468,
       "n": 10,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "v1",
       "label": "Optimizer v1 (pack v2.6, start)",
       "x": null,
       "y": 3.5099,
       "ci_low": 3.0156,
       "ci_high": 4.028,
       "n": 10,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "v2",
       "label": "Optimizer v2, rejected",
       "x": null,
       "y": 3.4436,
       "ci_low": 3.0942,
       "ci_high": 3.8072,
       "n": 10,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "v3",
       "label": "Optimizer v3, kept",
       "x": null,
       "y": 3.5616,
       "ci_low": 3.3082,
       "ci_high": 3.8615,
       "n": 10,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "v4",
       "label": "Optimizer v4, kept, best",
       "x": null,
       "y": 3.6447,
       "ci_low": 3.286,
       "ci_high": 4.0682,
       "n": 10,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Judge pascal-1 is also the policy model. Selection on held-out.",
   "status": "ready",
   "source": [
    "evals/optimizer/runs.csv",
    "evals/optimizer/versions.csv",
    "evals/runs/2026-10-31-measured/results.csv"
   ],
   "n": 10,
   "judge": "pascal-1",
   "date": "2026-10-31"
  },
  {
   "id": "judge_agreement",
   "figure": 8,
   "title": "Judge agreement",
   "subtitle": "pascal-1 vs gemini-4-argon",
   "headline": {
    "value": 0.308,
    "label": "Pearson r",
    "n": 42
   },
   "kind": "scatter",
   "png": "charts/08-judge-agreement.png",
   "svg": "charts/08-judge-agreement.svg",
   "series": [
    {
     "id": "pairs",
     "label": "outputs",
     "points": [
      {
       "key": "pascal-1|CG1",
       "label": "pascal-1 CG1",
       "x": 2.0,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6-astra|CG1",
       "label": "gpt-6-astra CG1",
       "x": 3.143,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra|CG1",
       "label": "gpt-6.1-terra CG1",
       "x": 2.571,
       "y": 3.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1|CG3",
       "label": "pascal-1 CG3",
       "x": 2.857,
       "y": 4.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6-astra|FG-L1",
       "label": "gpt-6-astra FG-L1",
       "x": 3.571,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra|CG2",
       "label": "gpt-6.1-terra CG2",
       "x": 2.429,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1|FG-L1",
       "label": "pascal-1 FG-L1",
       "x": 4.0,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1|CG4",
       "label": "pascal-1 CG4",
       "x": 4.143,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1|CG2",
       "label": "pascal-1 CG2",
       "x": 2.571,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra|CG4",
       "label": "gpt-6.1-terra CG4",
       "x": 3.429,
       "y": 4.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6-astra|CG2",
       "label": "gpt-6-astra CG2",
       "x": 3.429,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6-astra|CG4",
       "label": "gpt-6-astra CG4",
       "x": 4.0,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6-astra|CG3",
       "label": "gpt-6-astra CG3",
       "x": 2.857,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra|CG3",
       "label": "gpt-6.1-terra CG3",
       "x": 2.857,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback|CG1",
       "label": "pascal-1-fallback CG1",
       "x": 2.857,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1|01",
       "label": "pascal-1 01",
       "x": 3.429,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra|FG-L1",
       "label": "gpt-6.1-terra FG-L1",
       "x": 3.429,
       "y": 4.429,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6-astra|01",
       "label": "gpt-6-astra 01",
       "x": 3.286,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback|CG4",
       "label": "pascal-1-fallback CG4",
       "x": 3.286,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon|CG1",
       "label": "gemini-4-argon CG1",
       "x": 2.0,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback|CG2",
       "label": "pascal-1-fallback CG2",
       "x": 3.286,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback|CG3",
       "label": "pascal-1-fallback CG3",
       "x": 3.286,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback|FG-L1",
       "label": "pascal-1-fallback FG-L1",
       "x": 3.714,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "pascal-1-fallback|01",
       "label": "pascal-1-fallback 01",
       "x": 3.143,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon|CG4",
       "label": "gemini-4-argon CG4",
       "x": 3.286,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon|CG2",
       "label": "gemini-4-argon CG2",
       "x": 2.714,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7|CG1",
       "label": "grok-4.7 CG1",
       "x": 2.0,
       "y": 3.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon|CG3",
       "label": "gemini-4-argon CG3",
       "x": 2.286,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon|01",
       "label": "gemini-4-argon 01",
       "x": 2.429,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "google/gemini-4-argon|FG-L1",
       "label": "gemini-4-argon FG-L1",
       "x": 3.429,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5|CG1",
       "label": "claude-haiku-5.5 CG1",
       "x": 2.0,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5|CG2",
       "label": "claude-haiku-5.5 CG2",
       "x": 2.0,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5|CG3",
       "label": "claude-haiku-5.5 CG3",
       "x": 2.143,
       "y": 4.571,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5|CG4",
       "label": "claude-haiku-5.5 CG4",
       "x": 3.0,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7|CG2",
       "label": "grok-4.7 CG2",
       "x": 2.857,
       "y": 4.857,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7|CG4",
       "label": "grok-4.7 CG4",
       "x": 3.429,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7|CG3",
       "label": "grok-4.7 CG3",
       "x": 3.0,
       "y": 4.429,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5|FG-L1",
       "label": "claude-haiku-5.5 FG-L1",
       "x": 3.429,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7|FG-L1",
       "label": "grok-4.7 FG-L1",
       "x": 3.286,
       "y": 4.714,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "anthropic/claude-haiku-5.5|01",
       "label": "claude-haiku-5.5 01",
       "x": 2.0,
       "y": 4.429,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "x-ai/grok-4.7|01",
       "label": "grok-4.7 01",
       "x": 2.857,
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "openai/gpt-6.1-terra|01",
       "label": "gpt-6.1-terra 01",
       "x": 3.0,
       "y": 4.286,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "rank_second_judge",
     "label": "weighted score by second judge",
     "points": [
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 4.9523,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 2.5716
       }
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 4.9285,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 3.1906
       }
      },
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 4.8649,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 3.3493
       }
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 4.8254,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 2.9762
       }
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 4.738,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 2.2699
       }
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 4.4286,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 2.8096
       }
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 4.2938,
       "ci_low": null,
       "ci_high": null,
       "n": 6,
       "basis": "measured",
       "extra": {
        "primary_judge": 2.8493
       }
      }
     ]
    }
   ],
   "footnote": "Agreement with lawyers not yet measured.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-measured/xjudge.csv"
   ],
   "n": 42,
   "judge": "pascal-1; gemini-4-argon",
   "date": "2026-10-31"
  },
  {
   "id": "intent_routing",
   "figure": 9,
   "title": "Intent and routing",
   "subtitle": "regex router vs LLM intent step, same cases",
   "headline": {
    "value": 0.9381,
    "label": "LLM intent step, task right",
    "n": 97
   },
   "kind": "routing",
   "png": "charts/09-intent-routing.png",
   "svg": "charts/09-intent-routing.svg",
   "series": [
    {
     "id": "regex",
     "label": "regex router (measured)",
     "points": [
      {
       "key": "agent_ok",
       "label": "task or path right",
       "x": null,
       "y": 0.4639,
       "ci_low": 0.3679,
       "ci_high": 0.5626,
       "n": 97,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "instrument_ok",
       "label": "document type right",
       "x": null,
       "y": 0.7216,
       "ci_low": 0.6253,
       "ci_high": 0.8011,
       "n": 97,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "amounts_ok",
       "label": "no money read from durations",
       "x": null,
       "y": 0.6495,
       "ci_low": 0.5505,
       "ci_high": 0.7371,
       "n": 97,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "llm_intent",
     "label": "LLM intent step",
     "points": [
      {
       "key": "task",
       "label": "task or path right",
       "x": null,
       "y": 0.9381,
       "ci_low": 0.8716,
       "ci_high": 0.9713,
       "n": 97,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "doc_type",
       "label": "document type right",
       "x": null,
       "y": 0.9485,
       "ci_low": 0.885,
       "ci_high": 0.9778,
       "n": 97,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "no_dur_money",
       "label": "no money read from durations",
       "x": null,
       "y": 1.0,
       "ci_low": 0.9619,
       "ci_high": 1.0,
       "n": 97,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "strict_pass",
     "label": "strict pass",
     "points": [
      {
       "key": "regex",
       "label": "regex router, all three",
       "x": null,
       "y": 0.2268,
       "ci_low": 0.1548,
       "ci_high": 0.3196,
       "n": 97,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "llm",
       "label": "LLM intent step, every field",
       "x": null,
       "y": 0.4948,
       "ci_low": 0.3975,
       "ci_high": 0.5926,
       "n": 97,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "drafting",
     "label": "drafting pass",
     "points": [
      {
       "key": "main_pascal-1",
       "label": "production prompt, pascal-1",
       "x": null,
       "y": 0.0,
       "ci_low": 0.0,
       "ci_high": 0.1246,
       "n": 27,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "main_astra",
       "label": "production prompt, gpt-6-astra",
       "x": null,
       "y": 0.0,
       "ci_low": 0.0,
       "ci_high": 0.4899,
       "n": 4,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "eng_pascal-1",
       "label": "request passed through, pascal-1",
       "x": null,
       "y": 0.7,
       "ci_low": 0.3968,
       "ci_high": 0.8922,
       "n": 10,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "fix_pascal-1",
       "label": "proposed fix v1, pascal-1",
       "x": null,
       "y": 0.963,
       "ci_low": 0.8172,
       "ci_high": 0.9934,
       "n": 27,
       "basis": "measured",
       "extra": null
      }
     ]
    },
    {
     "id": "lint",
     "label": "prompt lint failures",
     "points": [
      {
       "key": "before",
       "label": "before hardening",
       "x": null,
       "y": 71.0,
       "ci_low": null,
       "ci_high": null,
       "n": 55,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "after",
       "label": "after hardening (proposal, not applied)",
       "x": null,
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": 55,
       "basis": "measured",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Deterministic scorers; strict pass definitions differ between the two arms.",
   "status": "ready",
   "source": [
    "evals/routing/runs/route-only-v2-97.json",
    "evals/intent/runs/harden-c1-before.scores.jsonl",
    "evals/routing/runs/",
    "evals/proposed/hardening/lint-before.json",
    "evals/proposed/hardening/lint-after.json"
   ],
   "n": 97,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "self_serve",
   "figure": 10,
   "title": "Self-serve rate and cost avoided",
   "subtitle": "measured vs estimated",
   "headline": {
    "value": 0.0612,
    "label": "self-serve, measured",
    "n": 49
   },
   "kind": "selfserve",
   "png": "charts/10-self-serve.png",
   "svg": "charts/10-self-serve.svg",
   "series": [
    {
     "id": "rate",
     "label": "self-serve rate",
     "points": [
      {
       "key": "measured",
       "label": "measured",
       "x": null,
       "y": 0.0612,
       "ci_low": 0.021,
       "ci_high": 0.1652,
       "n": 49,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "estimated",
       "label": "estimate, not re-run",
       "x": null,
       "y": 0.7931,
       "ci_low": null,
       "ci_high": null,
       "n": 58,
       "basis": "estimated",
       "extra": null
      }
     ]
    },
    {
     "id": "cost_saved",
     "label": "lawyer cost avoided, USD",
     "points": [
      {
       "key": "measured",
       "label": "measured",
       "x": null,
       "y": 385.72,
       "ci_low": null,
       "ci_high": null,
       "n": 49,
       "basis": "measured",
       "extra": null
      },
      {
       "key": "estimated",
       "label": "estimate, upper bound",
       "x": null,
       "y": 13952.7,
       "ci_low": null,
       "ci_high": null,
       "n": 58,
       "basis": "estimated",
       "extra": null
      }
     ]
    }
   ],
   "footnote": "Estimates are labelled and drawn hatched.",
   "status": "ready",
   "source": [
    "evals/selfserve/self-serve.json"
   ],
   "n": 49,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "live_run",
   "figure": 11,
   "title": "Live staging run",
   "subtitle": "18 tasks",
   "headline": null,
   "kind": "bar_ci",
   "png": "charts/11-live-run-slot.png",
   "svg": "charts/11-live-run-slot.svg",
   "series": [
    {
     "id": "live",
     "label": "staging",
     "points": []
    }
   ],
   "footnote": "Not yet measured. Fills when the staging results are on disk.",
   "status": "pending",
   "source": [],
   "n": null,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "environment_loop",
   "figure": 12,
   "title": "Contract drafting as an RL environment",
   "subtitle": "state, action, reward, adversary, transition, terminal, gate",
   "headline": null,
   "kind": "diagram",
   "png": "charts/12-environment-loop.png",
   "svg": "charts/12-environment-loop.svg",
   "series": [
    {
     "id": "card",
     "label": "environment card",
     "points": [
      {
       "key": "state",
       "label": "state",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "not_run",
       "extra": {
        "text": "request + resolved entity + Vault context"
       }
      },
      {
       "key": "action",
       "label": "action",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "not_run",
       "extra": {
        "text": "the draft"
       }
      },
      {
       "key": "reward",
       "label": "reward",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "not_run",
       "extra": {
        "text": "rubric v3 task score with hard-fail cap 1.5 (realism rubric 0 to 5 measured so far)"
       }
      },
      {
       "key": "adversary",
       "label": "adversary",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "text": "jurisdiction panel: senior partner + associate, P0 to P3 findings"
       }
      },
      {
       "key": "transition",
       "label": "transition",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "text": "repair: drafter gets its own draft plus findings"
       }
      },
      {
       "key": "terminal",
       "label": "terminal",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "text": "0 P0, 0 P1, fewer than 2 P2, or 4 reviews"
       }
      },
      {
       "key": "gate",
       "label": "gate",
       "x": null,
       "y": null,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "not_run",
       "extra": {
        "text": "prompt patch kept only if total score holds and no case gains a P0 or P1"
       }
      }
     ]
    }
   ],
   "footnote": "Rubric v3 reward and prompt-patch gate not yet run live.",
   "status": "ready",
   "source": [
    "evals/proposed/contract-hillclimb/loop.md",
    "evals/proposed/contract-hillclimb/0001-contract-hillclimb.patch"
   ],
   "n": 12,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "environment_learning_curve",
   "figure": 13,
   "title": "Mean reward by round",
   "subtitle": "12 episodes on main",
   "headline": {
    "value": 4.4042,
    "label": "mean reward at R2",
    "n": 12
   },
   "kind": "trajectory",
   "png": "charts/13-episode-learning-curve.png",
   "svg": "charts/13-episode-learning-curve.svg",
   "series": [
    {
     "id": "mean",
     "label": "mean reward",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 2.3333,
       "ci_low": 1.7158,
       "ci_high": 2.9158,
       "n": 12,
       "basis": "measured",
       "extra": {
        "P0": 19,
        "P1": 33,
        "P2": 23,
        "P3": 2,
        "shippable": 0
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.3325,
       "ci_low": 3.8933,
       "ci_high": 4.64,
       "n": 12,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 3,
        "P2": 7,
        "P3": 9,
        "shippable": 10
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.4042,
       "ci_low": 3.9525,
       "ci_high": 4.7242,
       "n": 12,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 2,
        "P2": 4,
        "P3": 10,
        "shippable": 11
       }
      }
     ]
    },
    {
     "id": "S01",
     "label": "S01",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 0.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S02",
     "label": "S02",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 0.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S03",
     "label": "S03",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S04",
     "label": "S04",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.14,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.57,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.57,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S05",
     "label": "S05",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S06",
     "label": "S06",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.57,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      }
     ]
    },
    {
     "id": "S07",
     "label": "S07",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.14,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      }
     ]
    },
    {
     "id": "S08",
     "label": "S08",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 1.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S09",
     "label": "S09",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 1.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      }
     ]
    },
    {
     "id": "S10",
     "label": "S10",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 1.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S11",
     "label": "S11",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 2.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    },
    {
     "id": "S12",
     "label": "S12",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 2.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 2.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 2.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "carried": true
       }
      }
     ]
    }
   ],
   "footnote": "Reviewer-graded; repairs by the same reviewer. Last round carries forward.",
   "status": "ready",
   "source": [
    "evals/contract-realism.md",
    "evals/contract-realism/findings/summary.json",
    "evals/contract-realism/findings/round-0.json",
    "evals/contract-realism/findings/round-1.json",
    "evals/contract-realism/findings/round-2.json",
    "evals/contract-realism/findings_data.py"
   ],
   "n": 12,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "environment_episodes",
   "figure": 14,
   "title": "Episode traces",
   "subtitle": "12 episodes on main",
   "headline": null,
   "kind": "small_multiples",
   "png": "charts/14-episode-small-multiples.png",
   "svg": "charts/14-episode-small-multiples.svg",
   "series": [
    {
     "id": "S01",
     "label": "S01 Mutual NDA",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 0.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 3,
        "P1": 3,
        "P2": 2,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S02",
     "label": "S02 NDA and MFN side letter",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 0.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 3,
        "P1": 1,
        "P2": 1,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S03",
     "label": "S03 NDA",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 4,
        "P2": 3,
        "P3": 1,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S04",
     "label": "S04 NDA",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.14,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 2,
        "P2": 3,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.57,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S05",
     "label": "S05 Services agreement",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 3,
        "P2": 3,
        "P3": 1,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S06",
     "label": "S06 Services agreement",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 2,
        "P2": 3,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.57,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S07",
     "label": "S07 Services agreement",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 3.14,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 3,
        "P2": 2,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 2,
        "P3": 0,
        "shippable": true,
        "clean": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 0,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S08",
     "label": "S08 Consulting agreement",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 1.43,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 2,
        "P1": 5,
        "P2": 1,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S09",
     "label": "S09 SAFE with MFN",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 1.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 4,
        "P2": 1,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 1,
        "P2": 1,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R2",
       "label": "R2",
       "x": 2,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S10",
     "label": "S10 SAFE, post-money template",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 1.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 2,
        "P1": 2,
        "P2": 2,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 4.86,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S11",
     "label": "S11 Jurisdiction law blocks",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 2.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 5,
        "P1": 2,
        "P2": 1,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      },
      {
       "key": "R1",
       "label": "R1",
       "x": 1,
       "y": 3.71,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1,
        "shippable": true,
        "clean": true
       }
      }
     ]
    },
    {
     "id": "S12",
     "label": "S12 Demo contract set",
     "points": [
      {
       "key": "R0",
       "label": "R0",
       "x": 0,
       "y": 2.29,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 2,
        "P2": 1,
        "P3": 0,
        "shippable": false,
        "clean": false
       }
      }
     ]
    }
   ],
   "footnote": "Reviewer-graded.",
   "status": "ready",
   "source": [
    "evals/contract-realism.md",
    "evals/contract-realism/findings/summary.json",
    "evals/contract-realism/findings/round-0.json",
    "evals/contract-realism/findings/round-1.json",
    "evals/contract-realism/findings/round-2.json",
    "evals/contract-realism/findings_data.py"
   ],
   "n": 12,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "environment_findings_heatmap",
   "figure": 15,
   "title": "Findings by category and round",
   "subtitle": "12 main samples + 8 branch fixtures",
   "headline": null,
   "kind": "heatmap",
   "png": "charts/15-findings-heatmap.png",
   "svg": "charts/15-findings-heatmap.svg",
   "series": [
    {
     "id": "r0",
     "label": "R0",
     "points": [
      {
       "key": "unenforceable",
       "label": "Unenforceable",
       "x": "R0",
       "y": 3.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 2,
        "P2": 1,
        "P3": 0
       }
      },
      {
       "key": "wrong_jurisdiction",
       "label": "Wrong jurisdiction",
       "x": "R0",
       "y": 12.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 2,
        "P1": 3,
        "P2": 7,
        "P3": 0
       }
      },
      {
       "key": "missing_mandatory_term",
       "label": "Missing mandatory term",
       "x": "R0",
       "y": 28.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 6,
        "P1": 17,
        "P2": 5,
        "P3": 0
       }
      },
      {
       "key": "defined_term_drift",
       "label": "Defined term drift",
       "x": "R0",
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 2,
        "P2": 2,
        "P3": 0
       }
      },
      {
       "key": "placeholder",
       "label": "Placeholder",
       "x": "R0",
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 3,
        "P1": 1,
        "P2": 1,
        "P3": 0
       }
      },
      {
       "key": "commercial_nonsense",
       "label": "Commercial nonsense",
       "x": "R0",
       "y": 7.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 2,
        "P1": 5,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "ai_tell",
       "label": "AI tell",
       "x": "R0",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "house_style",
       "label": "House style",
       "x": "R0",
       "y": 9.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 7,
        "P3": 2
       }
      },
      {
       "key": "personalisation",
       "label": "Personalisation",
       "x": "R0",
       "y": 2.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 2,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "stale_law",
       "label": "Stale law",
       "x": "R0",
       "y": 6.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 5,
        "P1": 1,
        "P2": 0,
        "P3": 0
       }
      }
     ]
    },
    {
     "id": "r1",
     "label": "R1",
     "points": [
      {
       "key": "unenforceable",
       "label": "Unenforceable",
       "x": "R1",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 0
       }
      },
      {
       "key": "wrong_jurisdiction",
       "label": "Wrong jurisdiction",
       "x": "R1",
       "y": 2.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 1
       }
      },
      {
       "key": "missing_mandatory_term",
       "label": "Missing mandatory term",
       "x": "R1",
       "y": 2.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 1
       }
      },
      {
       "key": "defined_term_drift",
       "label": "Defined term drift",
       "x": "R1",
       "y": 3.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 1,
        "P2": 1,
        "P3": 1
       }
      },
      {
       "key": "placeholder",
       "label": "Placeholder",
       "x": "R1",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "commercial_nonsense",
       "label": "Commercial nonsense",
       "x": "R1",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "ai_tell",
       "label": "AI tell",
       "x": "R1",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "house_style",
       "label": "House style",
       "x": "R1",
       "y": 5.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 2,
        "P3": 3
       }
      },
      {
       "key": "personalisation",
       "label": "Personalisation",
       "x": "R1",
       "y": 2.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 2
       }
      },
      {
       "key": "stale_law",
       "label": "Stale law",
       "x": "R1",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1
       }
      }
     ]
    },
    {
     "id": "r2",
     "label": "R2",
     "points": [
      {
       "key": "unenforceable",
       "label": "Unenforceable",
       "x": "R2",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 1,
        "P3": 0
       }
      },
      {
       "key": "wrong_jurisdiction",
       "label": "Wrong jurisdiction",
       "x": "R2",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1
       }
      },
      {
       "key": "missing_mandatory_term",
       "label": "Missing mandatory term",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "defined_term_drift",
       "label": "Defined term drift",
       "x": "R2",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 1
       }
      },
      {
       "key": "placeholder",
       "label": "Placeholder",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "commercial_nonsense",
       "label": "Commercial nonsense",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "ai_tell",
       "label": "AI tell",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "house_style",
       "label": "House style",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "personalisation",
       "label": "Personalisation",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "stale_law",
       "label": "Stale law",
       "x": "R2",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      }
     ]
    },
    {
     "id": "branch_r0",
     "label": "Branch R0",
     "points": [
      {
       "key": "unenforceable",
       "label": "Unenforceable",
       "x": "Branch R0",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "wrong_jurisdiction",
       "label": "Wrong jurisdiction",
       "x": "Branch R0",
       "y": 2.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 2,
        "P3": 0
       }
      },
      {
       "key": "missing_mandatory_term",
       "label": "Missing mandatory term",
       "x": "Branch R0",
       "y": 13.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 8,
        "P2": 3,
        "P3": 1
       }
      },
      {
       "key": "defined_term_drift",
       "label": "Defined term drift",
       "x": "Branch R0",
       "y": 3.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 1,
        "P2": 1,
        "P3": 1
       }
      },
      {
       "key": "placeholder",
       "label": "Placeholder",
       "x": "Branch R0",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "commercial_nonsense",
       "label": "Commercial nonsense",
       "x": "Branch R0",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "ai_tell",
       "label": "AI tell",
       "x": "Branch R0",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "house_style",
       "label": "House style",
       "x": "Branch R0",
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 1,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "personalisation",
       "label": "Personalisation",
       "x": "Branch R0",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "stale_law",
       "label": "Stale law",
       "x": "Branch R0",
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": null,
       "basis": "measured",
       "extra": {
        "P0": 0,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      }
     ]
    }
   ],
   "footnote": "Categories keyword-derived.",
   "status": "ready",
   "source": [
    "evals/contract-realism.md",
    "evals/contract-realism/findings/summary.json",
    "evals/contract-realism/findings/round-0.json",
    "evals/contract-realism/findings/round-1.json",
    "evals/contract-realism/findings/round-2.json",
    "evals/contract-realism/findings_data.py"
   ],
   "n": 117,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "environment_policy_updates",
   "figure": 16,
   "title": "Policy updates",
   "subtitle": "top generalised rules by weighted severity",
   "headline": null,
   "kind": "bar",
   "png": "charts/16-policy-updates.png",
   "svg": "charts/16-policy-updates.svg",
   "series": [
    {
     "id": "rules",
     "label": "rule clusters",
     "points": [
      {
       "key": "rule1",
       "label": "Parties block completeness is a precondition, not an optional field.",
       "x": null,
       "y": 80.0,
       "ci_low": null,
       "ci_high": null,
       "n": 8,
       "basis": "measured",
       "extra": {
        "category": "missing_mandatory_term",
        "patches": "P-01 P-02 P-13 P-14 P-15",
        "P0": 0,
        "P1": 8,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule2",
       "label": "Prompts cite current law only; a law-change table drives both prompts and engine.",
       "x": null,
       "y": 75.0,
       "ci_low": null,
       "ci_high": null,
       "n": 3,
       "basis": "measured",
       "extra": {
        "category": "stale_law",
        "patches": "P-12 P-15",
        "P0": 3,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule3",
       "label": "Every contract sample has a governing law and a jurisdiction-correct forum clause.",
       "x": null,
       "y": 50.0,
       "ci_low": null,
       "ci_high": null,
       "n": 2,
       "basis": "measured",
       "extra": {
        "category": "missing_mandatory_term",
        "patches": "P-09 P-10 P-15",
        "P0": 2,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule4",
       "label": "Seeds and fixtures are realistic documents, not mail-merge shells.",
       "x": null,
       "y": 50.0,
       "ci_low": null,
       "ci_high": null,
       "n": 2,
       "basis": "measured",
       "extra": {
        "category": "placeholder",
        "patches": "P-13 P-15",
        "P0": 2,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule5",
       "label": "A document must contain the operative obligation its title promises; an NDA without a confidentiality covenant fails the eval.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "missing_mandatory_term",
        "patches": "P-09",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule6",
       "label": "A sample presented as signed is complete.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "missing_mandatory_term",
        "patches": "P-13",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule7",
       "label": "Entity form must match the legal suffix and the register; never fall back to 'a corporation'.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "wrong_jurisdiction",
        "patches": "P-01",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule8",
       "label": "Every cross-referenced schedule exists.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "commercial_nonsense",
        "patches": "P-13",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule9",
       "label": "Every economic term resolves to defined inputs.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "commercial_nonsense",
        "patches": "P-13",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule10",
       "label": "Every party is described with full legal name, entity type, jurisdiction of incorporation, company number where customary and registered or principal office.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "missing_mandatory_term",
        "patches": "P-09 P-15",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule11",
       "label": "Examples in prompts are current and verified; models copy them.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "stale_law",
        "patches": "P-12 P-15",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      },
      {
       "key": "rule12",
       "label": "Fixtures that render a signable document reuse a vetted sample constant; no inline stubs.",
       "x": null,
       "y": 25.0,
       "ci_low": null,
       "ci_high": null,
       "n": 1,
       "basis": "measured",
       "extra": {
        "category": "placeholder",
        "patches": "P-10 P-15",
        "P0": 1,
        "P1": 0,
        "P2": 0,
        "P3": 0
       }
      }
     ]
    }
   ],
   "footnote": "Patches proposed, not applied.",
   "status": "ready",
   "source": [
    "evals/contract-realism.md",
    "evals/contract-realism/findings/summary.json",
    "evals/contract-realism/findings/round-0.json",
    "evals/contract-realism/findings/round-1.json",
    "evals/contract-realism/findings/round-2.json",
    "evals/contract-realism/findings_data.py",
    "evals/proposed/contract-realism-patches.md"
   ],
   "n": 117,
   "judge": null,
   "date": "2026-10-31"
  },
  {
   "id": "environment_reward_gate",
   "figure": 17,
   "title": "Rubric v3 pre-check gate",
   "subtitle": "141 of 221 auto-fail",
   "headline": {
    "value": 0.638,
    "label": "auto-fail share",
    "n": 221
   },
   "kind": "bar_ci",
   "png": "charts/17-v3-precheck-gate.png",
   "svg": "charts/17-v3-precheck-gate.svg",
   "series": [
    {
     "id": "by_reason",
     "label": "by reason",
     "points": [
      {
       "key": "over_budget",
       "label": "over_budget",
       "x": null,
       "y": 122.0,
       "ci_low": null,
       "ci_high": null,
       "n": 221,
       "basis": "measured",
       "extra": {
        "only_reason": 112
       }
      },
      {
       "key": "no_citations",
       "label": "no_citations",
       "x": null,
       "y": 29.0,
       "ci_low": null,
       "ci_high": null,
       "n": 221,
       "basis": "measured",
       "extra": {
        "only_reason": 17
       }
      },
      {
       "key": "em_dash",
       "label": "em_dash",
       "x": null,
       "y": 2.0,
       "ci_low": null,
       "ci_high": null,
       "n": 221,
       "basis": "measured",
       "extra": {
        "only_reason": 0
       }
      },
      {
       "key": "placeholder",
       "label": "placeholder",
       "x": null,
       "y": 1.0,
       "ci_low": null,
       "ci_high": null,
       "n": 221,
       "basis": "measured",
       "extra": {
        "only_reason": 0
       }
      },
      {
       "key": "party_mismatch",
       "label": "party_mismatch",
       "x": null,
       "y": 0.0,
       "ci_low": null,
       "ci_high": null,
       "n": 221,
       "basis": "measured",
       "extra": {
        "only_reason": 0
       }
      }
     ]
    },
    {
     "id": "by_prompt",
     "label": "by prompt version",
     "points": [
      {
       "key": "production app",
       "label": "production app",
       "x": null,
       "y": 0.9474,
       "ci_low": 0.7536,
       "ci_high": 0.9906,
       "n": 19,
       "basis": "measured",
       "extra": {
        "auto_fail": 18
       }
      },
      {
       "key": "optimizer v1 (pack v2.6 s0-7)",
       "label": "optimizer v1 (pack v2.6 s0-7)",
       "x": null,
       "y": 0.8,
       "ci_low": 0.3755,
       "ci_high": 0.9638,
       "n": 5,
       "basis": "measured",
       "extra": {
        "auto_fail": 4
       }
      },
      {
       "key": "production QUALITY_BAR",
       "label": "production QUALITY_BAR",
       "x": null,
       "y": 0.7647,
       "ci_low": 0.5274,
       "ci_high": 0.9045,
       "n": 17,
       "basis": "measured",
       "extra": {
        "auto_fail": 13
       }
      },
      {
       "key": "pack v2.6",
       "label": "pack v2.6",
       "x": null,
       "y": 0.7576,
       "ci_low": 0.6419,
       "ci_high": 0.8449,
       "n": 66,
       "basis": "measured",
       "extra": {
        "auto_fail": 50
       }
      },
      {
       "key": "optimizer v1",
       "label": "optimizer v1",
       "x": null,
       "y": 0.5263,
       "ci_low": 0.3726,
       "ci_high": 0.6752,
       "n": 38,
       "basis": "measured",
       "extra": {
        "auto_fail": 20
       }
      },
      {
       "key": "optimizer v2",
       "label": "optimizer v2",
       "x": null,
       "y": 0.5263,
       "ci_low": 0.3726,
       "ci_high": 0.6752,
       "n": 38,
       "basis": "measured",
       "extra": {
        "auto_fail": 20
       }
      },
      {
       "key": "optimizer v3",
       "label": "optimizer v3",
       "x": null,
       "y": 0.4211,
       "ci_low": 0.2314,
       "ci_high": 0.6372,
       "n": 19,
       "basis": "measured",
       "extra": {
        "auto_fail": 8
       }
      },
      {
       "key": "optimizer v4",
       "label": "optimizer v4",
       "x": null,
       "y": 0.4211,
       "ci_low": 0.2314,
       "ci_high": 0.6372,
       "n": 19,
       "basis": "measured",
       "extra": {
        "auto_fail": 8
       }
      }
     ]
    },
    {
     "id": "by_model",
     "label": "by model",
     "points": [
      {
       "key": "google/gemini-4-argon",
       "label": "gemini-4-argon",
       "x": null,
       "y": 1.0,
       "ci_low": 0.7225,
       "ci_high": 1.0,
       "n": 10,
       "basis": "measured",
       "extra": {
        "auto_fail": 10
       }
      },
      {
       "key": "not recorded",
       "label": "not recorded",
       "x": null,
       "y": 0.9474,
       "ci_low": 0.7536,
       "ci_high": 0.9906,
       "n": 19,
       "basis": "measured",
       "extra": {
        "auto_fail": 18
       }
      },
      {
       "key": "x-ai/grok-4.7",
       "label": "grok-4.7",
       "x": null,
       "y": 0.9,
       "ci_low": 0.5958,
       "ci_high": 0.9821,
       "n": 10,
       "basis": "measured",
       "extra": {
        "auto_fail": 9
       }
      },
      {
       "key": "openai/gpt-6-astra",
       "label": "gpt-6-astra",
       "x": null,
       "y": 0.8333,
       "ci_low": 0.552,
       "ci_high": 0.953,
       "n": 12,
       "basis": "measured",
       "extra": {
        "auto_fail": 10
       }
      },
      {
       "key": "anthropic/claude-haiku-5.5",
       "label": "claude-haiku-5.5",
       "x": null,
       "y": 0.8,
       "ci_low": 0.4902,
       "ci_high": 0.9433,
       "n": 10,
       "basis": "measured",
       "extra": {
        "auto_fail": 8
       }
      },
      {
       "key": "pascal-1-fallback",
       "label": "pascal-1-fallback",
       "x": null,
       "y": 0.7,
       "ci_low": 0.3968,
       "ci_high": 0.8922,
       "n": 10,
       "basis": "measured",
       "extra": {
        "auto_fail": 7
       }
      },
      {
       "key": "pascal-1",
       "label": "pascal-1",
       "x": null,
       "y": 0.5286,
       "ci_low": 0.4462,
       "ci_high": 0.6094,
       "n": 140,
       "basis": "measured",
       "extra": {
        "auto_fail": 74
       }
      },
      {
       "key": "openai/gpt-6.1-terra",
       "label": "gpt-6.1-terra",
       "x": null,
       "y": 0.5,
       "ci_low": 0.2366,
       "ci_high": 0.7634,
       "n": 10,
       "basis": "measured",
       "extra": {
        "auto_fail": 5
       }
      }
     ]
    }
   ],
   "footnote": "Deterministic pre-check only; v3 judge scores not yet measured.",
   "status": "ready",
   "source": [
    "evals/runs/2026-10-31-v3-precheck/results.csv",
    "evals/runs/2026-10-31-v3-precheck/summary.md",
    "evals/RUBRIC-v3.md"
   ],
   "n": 221,
   "judge": null,
   "date": "2026-10-31"
  }
 ],
 "customers": [
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-a",
   "segment": "Regulated vendor data terms (customer-derived)",
   "filter": "/training?segment=seg-a",
   "depth": "deep",
   "rubric_weights": {
    "accuracy": 1.5,
    "grounding": 1.0,
    "specificity": 1.0,
    "actionability": 1.5,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGA-1",
     "title": "Regulated vendor data terms task 1",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGA-2",
     "title": "Regulated vendor data terms task 2",
     "weight": 2.5,
     "status": "not_run"
    },
    {
     "id": "SEGA-3",
     "title": "Regulated vendor data terms task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGA-4",
     "title": "Regulated vendor data terms task 4",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGA-5",
     "title": "Regulated vendor data terms task 5",
     "weight": 1.5,
     "status": "not_run"
    },
    {
     "id": "SEGA-6",
     "title": "Regulated vendor data terms task 6",
     "weight": 1.5,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 116
  },
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-b",
   "segment": "Sports sponsorship (customer-derived)",
   "filter": "/training?segment=seg-b",
   "depth": "deep",
   "rubric_weights": {
    "accuracy": 1.0,
    "grounding": 1.0,
    "specificity": 1.5,
    "actionability": 1.5,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGB-1",
     "title": "Sports sponsorship task 1",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGB-2",
     "title": "Sports sponsorship task 2",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGB-3",
     "title": "Sports sponsorship task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGB-4",
     "title": "Sports sponsorship task 4",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGB-5",
     "title": "Sports sponsorship task 5",
     "weight": 1.5,
     "status": "not_run"
    },
    {
     "id": "SEGB-6",
     "title": "Sports sponsorship task 6",
     "weight": 1.5,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 7
  },
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-c",
   "segment": "Investor outreach compliance (customer-derived)",
   "filter": "/training?segment=seg-c",
   "depth": "deep",
   "rubric_weights": {
    "accuracy": 1.5,
    "grounding": 1.5,
    "specificity": 1.0,
    "actionability": 1.0,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGC-1",
     "title": "Investor outreach compliance task 1",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGC-2",
     "title": "Investor outreach compliance task 2",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGC-3",
     "title": "Investor outreach compliance task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGC-4",
     "title": "Investor outreach compliance task 4",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGC-5",
     "title": "Investor outreach compliance task 5",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGC-6",
     "title": "Investor outreach compliance task 6",
     "weight": 1.5,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 6
  },
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-d",
   "segment": "SaaS services agreements (customer-derived)",
   "filter": "/training?segment=seg-d",
   "depth": "medium",
   "rubric_weights": {
    "accuracy": 1.5,
    "grounding": 1.0,
    "specificity": 1.5,
    "actionability": 1.0,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGD-1",
     "title": "SaaS services agreements task 1",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGD-2",
     "title": "SaaS services agreements task 2",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGD-3",
     "title": "SaaS services agreements task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGD-4",
     "title": "SaaS services agreements task 4",
     "weight": 1.5,
     "status": "not_run"
    },
    {
     "id": "SEGD-5",
     "title": "SaaS services agreements task 5",
     "weight": 1.0,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 5
  },
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-e",
   "segment": "Overnight MSA review (customer-derived)",
   "filter": "/training?segment=seg-e",
   "depth": "medium",
   "rubric_weights": {
    "accuracy": 1.0,
    "grounding": 1.0,
    "specificity": 1.5,
    "actionability": 1.5,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGE-1",
     "title": "Overnight MSA review task 1",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGE-2",
     "title": "Overnight MSA review task 2",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGE-3",
     "title": "Overnight MSA review task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGE-4",
     "title": "Overnight MSA review task 4",
     "weight": 1.5,
     "status": "not_run"
    },
    {
     "id": "SEGE-5",
     "title": "Overnight MSA review task 5",
     "weight": 1.0,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 5
  },
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-f",
   "segment": "Health data company set-up (customer-derived)",
   "filter": "/training?segment=seg-f",
   "depth": "short",
   "rubric_weights": {
    "accuracy": 1.5,
    "grounding": 1.0,
    "specificity": 1.0,
    "actionability": 1.5,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGF-1",
     "title": "Health data company set-up task 1",
     "weight": 2.5,
     "status": "not_run"
    },
    {
     "id": "SEGF-2",
     "title": "Health data company set-up task 2",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGF-3",
     "title": "Health data company set-up task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGF-4",
     "title": "Health data company set-up task 4",
     "weight": 1.0,
     "status": "not_run"
    },
    {
     "id": "SEGF-5",
     "title": "Health data company set-up task 5",
     "weight": 1.0,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 5
  },
  {
   "slug": null,
   "name": null,
   "segment_id": "seg-g",
   "segment": "Venture financing documents (customer-derived)",
   "filter": "/training?segment=seg-g",
   "depth": "short",
   "rubric_weights": {
    "accuracy": 1.0,
    "grounding": 1.0,
    "specificity": 1.5,
    "actionability": 1.5,
    "polish": 1.0,
    "e2e_wow": 2.0
   },
   "hard_fail_cap": 2.0,
   "tasks": [
    {
     "id": "SEGG-1",
     "title": "Venture financing documents task 1",
     "weight": 3.0,
     "status": "not_run"
    },
    {
     "id": "SEGG-2",
     "title": "Venture financing documents task 2",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGG-3",
     "title": "Venture financing documents task 3",
     "weight": 2.0,
     "status": "not_run"
    },
    {
     "id": "SEGG-4",
     "title": "Venture financing documents task 4",
     "weight": 1.0,
     "status": "not_run"
    },
    {
     "id": "SEGG-5",
     "title": "Venture financing documents task 5",
     "weight": 1.0,
     "status": "not_run"
    }
   ],
   "observations": [],
   "measured_overall": null,
   "rows": 5
  }
 ],
 "changelog": [
  {
   "version": "0.1",
   "date": "2026-10-31T05:46:00-07:00",
   "change": "Model bake-off: 7 models on Pascal pack v2.6, 10 tasks; citation check (438); second judge on 42 outputs.",
   "dataset_sha256": null,
   "results_sha256": null
  },
  {
   "version": "0.2",
   "date": "2026-10-31T06:25:00-07:00",
   "change": "Prompt optimizer v1 to v4 on pascal-1; 11-task held-out split (10 judged, R4 by leak scan).",
   "dataset_sha256": null,
   "results_sha256": null
  },
  {
   "version": "0.3",
   "date": "2026-10-31T11:45:00-07:00",
   "change": "Routing and drafting baseline ROUTE-C1; routing set grows to 97 cases (schema v2).",
   "dataset_sha256": null,
   "results_sha256": null
  },
  {
   "version": "0.4",
   "date": "2026-10-31T13:51:00-07:00",
   "change": "Static prompt lint on 55 prompts before and after hardening patches 01 to 10.",
   "dataset_sha256": null,
   "results_sha256": null
  },
  {
   "version": "1.0",
   "date": "2026-10-31T14:18:05-07:00",
   "change": "First PascalBench build: charts, data.json and grid rows regenerated from files on disk.",
   "dataset_sha256": "3c6f0b4e96ffd310ae5d9ed13337e1f9e69537951502524b3412cfb9f280a75a",
   "results_sha256": "703a6bb01233c7de9f03143ab7b77e1624b88f92aa22b13a7274221a9d05bbf7"
  },
  {
   "version": "1.1",
   "date": "2026-10-31T14:27:10-07:00",
   "change": "Public build is customer-name-free: data.public.json, rows.public.csv and docs show segments only; customer rows move to tenants/<slug>.json behind the access-code gate; leak check against the repo customer denylist runs before every write. Adds 50 RT tasks (not yet run) and the 102 RL routing cases to counts and grid rows. No scores changed.",
   "dataset_sha256": "d0ee22700cca7c8b849c8b5fcd6d874d8bd76a319264ba9338f47c2e12fad31e",
   "results_sha256": "3868b49334f2a787aff583b49fda6fd6a9e83d7295425c144019925ee319f31c"
  },
  {
   "version": "1.2",
   "date": "2026-10-31T15:17:11-07:00",
   "change": "Dataset changed (tasks, cases or rubric).",
   "dataset_sha256": "5a861b05eda31e3734b31265aa323e1a23a25c2d119580f176f7e341289f241d",
   "results_sha256": "3868b49334f2a787aff583b49fda6fd6a9e83d7295425c144019925ee319f31c"
  },
  {
   "version": "1.3",
   "date": "2026-10-31T15:52:14-07:00",
   "change": "Environment section: the contract hill-climb as an RL environment (state, action, reward, adversary, transition, terminal, gate). Charts 12 to 17 and data.environment from R-CONTRACT-REALISM (12 episodes on main: P0 19 to 0, P1 33 to 2, mean reward 2.33 to 4.40, 11 of 12 shippable), the rubric v3 pre-check (141 of 221 auto-fail) and the harness dry run. Live judge run not yet measured. No existing score changed.",
   "dataset_sha256": "5a861b05eda31e3734b31265aa323e1a23a25c2d119580f176f7e341289f241d",
   "results_sha256": "c587bd8eab240382a8a071db6fc45e8116c6bb7b5d88f7961930bf6b0e352f06"
  }
 ],
 "next_run": {
  "status": "planned",
  "target": "Pascal staging (qp971/pascal-1-vault)",
  "tasks": [
   {
    "id": "CG1",
    "category": "drafting",
    "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
    "jurisdiction": "DE",
    "weight": 4.0,
    "split": "train",
    "file": "evals/tasks/CG1-cold-mfn-waiver-upload.md"
   },
   {
    "id": "CG3",
    "category": "drafting",
    "title": "Cold: NDA change and send for signature (simulated)",
    "jurisdiction": "NY",
    "weight": 3.0,
    "split": "train",
    "file": "evals/tasks/CG3-cold-nda-change-and-sign.md"
   },
   {
    "id": "CG4",
    "category": "drafting",
    "title": "Cold: board consent for a SAFE round",
    "jurisdiction": "DE",
    "weight": 3.0,
    "split": "heldout",
    "file": "evals/tasks/CG4-cold-board-consent.md"
   },
   {
    "id": "01",
    "category": "drafting",
    "title": "MFN side letter waiver + cover email",
    "jurisdiction": "DE",
    "weight": 3.0,
    "split": "train",
    "file": "evals/tasks/01-mfn-side-letter-edit.md"
   },
   {
    "id": "22",
    "category": "drafting",
    "title": "Series A term sheet counter",
    "jurisdiction": "DE",
    "weight": 3.0,
    "split": "heldout",
    "file": "evals/tasks/22-series-a-term-sheet-review.md"
   },
   {
    "id": "03",
    "category": "redlining",
    "title": "DPA redline (customer-derived, Art. 28)",
    "jurisdiction": "EU",
    "weight": 2.0,
    "split": "heldout",
    "file": "evals/tasks/ (customer-derived file, held privately)"
   },
   {
    "id": "09",
    "category": "redlining",
    "title": "Mutual NDA redline",
    "jurisdiction": "CA;DE",
    "weight": 2.5,
    "split": "train",
    "file": "evals/tasks/09-mutual-nda-redline.md"
   },
   {
    "id": "11",
    "category": "redlining",
    "title": "Enterprise MSA counter-redline (NY)",
    "jurisdiction": "NY",
    "weight": 3.0,
    "split": "heldout",
    "file": "evals/tasks/11-saas-msa-customer-redline-ny.md"
   },
   {
    "id": "UC-M1",
    "category": "redlining",
    "title": "Counterparty markup: respond to a customer's markup of our DPA",
    "jurisdiction": "EU;EW",
    "weight": 2.5,
    "split": null,
    "file": "evals/tasks/UC-M1-markup-dpa-response.md"
   },
   {
    "id": "13",
    "category": "research",
    "title": "EU AI Act for AI sales platform",
    "jurisdiction": "EU",
    "weight": 2.0,
    "split": "train",
    "file": "evals/tasks/13-eu-ai-act-ai-sales-platform.md"
   },
   {
    "id": "18",
    "category": "research",
    "title": "California security deposit",
    "jurisdiction": "CA",
    "weight": 2.0,
    "split": "heldout",
    "file": "evals/tasks/18-california-security-deposit.md"
   },
   {
    "id": "28",
    "category": "research",
    "title": "CCPA access request clock and extension",
    "jurisdiction": "CA",
    "weight": 2.5,
    "split": null,
    "file": "evals/tasks/28-ccpa-access-request-clock.md"
   },
   {
    "id": "39",
    "category": "research",
    "title": "UK breach policy: 72 hours quote check",
    "jurisdiction": "EW",
    "weight": 1.5,
    "split": null,
    "file": "evals/tasks/39-uk-gdpr-breach-72-hours-quote-check.md"
   },
   {
    "id": "40",
    "category": "research",
    "title": "Citation check on a draft letter with a suspect case",
    "jurisdiction": "NY",
    "weight": 1.5,
    "split": null,
    "file": "evals/tasks/40-citation-check-brief-with-suspect-case.md"
   },
   {
    "id": "FG-L15",
    "category": "research",
    "title": "California work made for hire cite",
    "jurisdiction": "CA",
    "weight": 2.0,
    "split": null,
    "file": "evals/tasks/FG-L15-ca-work-for-hire-cite.md"
   },
   {
    "id": "RC-002",
    "category": "intent",
    "title": "send an MFN waiver to [email] for Harbor Lane Ventures to waive the MFN in",
    "jurisdiction": null,
    "weight": null,
    "split": null,
    "file": "evals/routing/cases.jsonl"
   },
   {
    "id": "IR-002",
    "category": "intent",
    "title": "We are about to raise our next round at Pellucid Systems, Inc. Please prepare an MFN waive",
    "jurisdiction": null,
    "weight": null,
    "split": null,
    "file": "evals/routing/cases.jsonl"
   },
   {
    "id": "RC-019",
    "category": "intent",
    "title": "Draft an independent contractor agreement for Falcon Grid Technologies LLC (Dubai mainland",
    "jurisdiction": "AE-ONSHORE",
    "weight": null,
    "split": null,
    "file": "evals/routing/cases.jsonl"
   }
  ],
  "results_path": "evals/runs/<date>-staging/results.csv"
 },
 "limitations": [
  "The primary judge (pascal-1) is also a contestant and the optimizer's policy model.",
  "Two LLM judges agree only weakly (r = 0.31); the second judge never scores lower and ranks its own model first. No lawyer labels yet.",
  "Small n: 6 common tasks per model. No model difference against gpt-6-astra survives Holm correction for 6 comparisons.",
  "One sample per task per model in the bake-off; sampling variance is not separated from task variance.",
  "gpt-6-astra ran 6 of 10 tasks to stay within budget, so its all-task score is not reported.",
  "The held-out split was also the acceptance metric across prompt versions, so the best version carries selection bias.",
  "The bake-off tests the prompt pack with one call, not Pascal's full pipeline (intake, routing, retrieval, signing).",
  "146 of 438 citations were model-checked rather than checked against fetched source text.",
  "Self-serve estimates are static reviews after queued fixes, never re-run; they are an upper bound.",
  "Customer pack tasks are targets and have not been run.",
  "Prompt lint results describe a proposed patch set that is not applied."
 ],
 "rows": [
  {
   "row_id": "bakeoff:v2.6:pascal-1:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 2.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:46:47-07:00",
   "cost_usd": 0.0491,
   "latency_s": 67.5,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/CG1.md",
   "notes": "Fails to preserve the source document’s defined terms: the replacement uses lowercase “waiver” instead of “Waiver” and omits the defined terms “SAFEs” and “MFN Provision,” substituting descriptive lan",
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6-astra:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:46:47-07:00",
   "cost_usd": 0.2861,
   "latency_s": 98.0,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6-astra/CG1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:46:57-07:00",
   "cost_usd": 0.0166,
   "latency_s": 26.9,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/CG1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:06:0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:33-07:00",
   "cost_usd": 0.0235,
   "latency_s": 47.5912,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/06.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 5.0,
   "task_score": 4.714,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:33-07:00",
   "cost_usd": 0.025,
   "latency_s": 45.1607,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:40-07:00",
   "cost_usd": 0.0287,
   "latency_s": 52.3135,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6-astra:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:45-07:00",
   "cost_usd": 0.1461,
   "latency_s": 59.6505,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6-astra/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.429,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:46-07:00",
   "cost_usd": 0.0178,
   "latency_s": 39.4418,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:50-07:00",
   "cost_usd": 0.0347,
   "latency_s": 63.8366,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:52-07:00",
   "cost_usd": 0.0336,
   "latency_s": 64.9104,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:47:55-07:00",
   "cost_usd": 0.0431,
   "latency_s": 68.2042,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:11-07:00",
   "cost_usd": 0.0126,
   "latency_s": 28.0865,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6-astra:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:14-07:00",
   "cost_usd": 0.2359,
   "latency_s": 88.5556,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6-astra/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6-astra:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:17-07:00",
   "cost_usd": 0.1974,
   "latency_s": 92.1159,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6-astra/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6-astra:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:22-07:00",
   "cost_usd": 0.238,
   "latency_s": 94.7241,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6-astra/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:22-07:00",
   "cost_usd": 0.0181,
   "latency_s": 40.1178,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:06:0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:23-07:00",
   "cost_usd": 0.0097,
   "latency_s": 23.143,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/06.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:26-07:00",
   "cost_usd": 0.0624,
   "latency_s": 100.6634,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.714,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:32-07:00",
   "cost_usd": 0.0126,
   "latency_s": 28.8183,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:13:0",
   "case_id": "13",
   "kind": "model_run",
   "title": "EU AI Act for AI sales platform",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 5.0,
   "task_score": 4.857,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:34-07:00",
   "cost_usd": 0.0534,
   "latency_s": 109.4205,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/13.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:36-07:00",
   "cost_usd": 0.175,
   "latency_s": 70.0,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/CG1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:13:0",
   "case_id": "13",
   "kind": "model_run",
   "title": "EU AI Act for AI sales platform",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:43-07:00",
   "cost_usd": 0.0174,
   "latency_s": 42.6922,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/13.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:48:46-07:00",
   "cost_usd": 0.0685,
   "latency_s": 118.5432,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1/01.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:11-07:00",
   "cost_usd": 0.0158,
   "latency_s": 35.4333,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6-astra:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:23-07:00",
   "cost_usd": 0.3865,
   "latency_s": 156.231,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6-astra/01.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:25-07:00",
   "cost_usd": 0.1191,
   "latency_s": 44.0567,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:25-07:00",
   "cost_usd": 0.0662,
   "latency_s": 75.7,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/CG1.md",
   "notes": "The purported redline replaces the source's defined-term formulation with generic 'Most Favored Nation' language rather than preserving 'MFN Provision'. It also substitutes a new Section 2 without sho",
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:34-07:00",
   "cost_usd": 0.15,
   "latency_s": 62.5525,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:35-07:00",
   "cost_usd": 0.1365,
   "latency_s": 53.5568,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:06:0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:40-07:00",
   "cost_usd": 0.1108,
   "latency_s": 46.7538,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/06.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.857,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:46-07:00",
   "cost_usd": 0.1043,
   "latency_s": 40.1724,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:13:0",
   "case_id": "13",
   "kind": "model_run",
   "title": "EU AI Act for AI sales platform",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:49:58-07:00",
   "cost_usd": 0.1432,
   "latency_s": 59.3295,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/13.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 3.714,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:14-07:00",
   "cost_usd": 0.1579,
   "latency_s": 61.2648,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:17-07:00",
   "cost_usd": 0.2163,
   "latency_s": 90.0583,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/01.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:pascal-1-fallback:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:29-07:00",
   "cost_usd": 0.2555,
   "latency_s": 103.7093,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/pascal-1-fallback/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:29-07:00",
   "cost_usd": 0.0386,
   "latency_s": 43.7229,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 2.714,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:39-07:00",
   "cost_usd": 0.0597,
   "latency_s": 62.535,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:41-07:00",
   "cost_usd": 0.0745,
   "latency_s": 127.7,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/CG1.md",
   "notes": "The replacement clean document drops the defined terms \"Waiver\" and \"MFN Provision\", using \"waiver\" and \"MFN\" instead. It also replaces the original consideration provision rather than preserving the ",
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.286,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:43-07:00",
   "cost_usd": 0.065,
   "latency_s": 69.3417,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:06:0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:52-07:00",
   "cost_usd": 0.0353,
   "latency_s": 52.5503,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/06.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:13:0",
   "case_id": "13",
   "kind": "model_run",
   "title": "EU AI Act for AI sales platform",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.429,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:50:57-07:00",
   "cost_usd": 0.0389,
   "latency_s": 46.5207,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/13.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:51:12-07:00",
   "cost_usd": 0.0786,
   "latency_s": 85.289,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/01.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:51:12-07:00",
   "cost_usd": 0.0401,
   "latency_s": 45.5584,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:51:13-07:00",
   "cost_usd": 0.0724,
   "latency_s": 81.6612,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "bakeoff:v2.6:gemini-4-argon:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:51:14-07:00",
   "cost_usd": 0.0452,
   "latency_s": 48.4365,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/google_gemini-4-argon/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:51:26-07:00",
   "cost_usd": 0.0046,
   "latency_s": 36.4,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/CG1.md",
   "notes": "The purported clean version replaces the document's defined term \"MFN Provision\" with \"MFN rights\" and omits defined-term operative language rather than preserving the user's document. This triggers t",
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:52:07-07:00",
   "cost_usd": 0.0039,
   "latency_s": 30.9845,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/CG2.md",
   "notes": "The redline changes the source defined term \"Subsequent Convertible Securities\" to \"Subsequent Convertible Security,\" violating the requirement to preserve defined terms.",
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:06:0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:52:37-07:00",
   "cost_usd": 0.0439,
   "latency_s": 91.9582,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/06.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 2.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.143,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:52:58-07:00",
   "cost_usd": 0.0048,
   "latency_s": 40.1632,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.0,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:53:21-07:00",
   "cost_usd": 0.0043,
   "latency_s": 34.1577,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 2.857,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:53:24-07:00",
   "cost_usd": 0.0794,
   "latency_s": 164.1117,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:06:0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:53:57-07:00",
   "cost_usd": 0.0031,
   "latency_s": 24.2724,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/06.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:54:09-07:00",
   "cost_usd": 0.0811,
   "latency_s": 196.6704,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:54:32-07:00",
   "cost_usd": 0.0076,
   "latency_s": 60.1901,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.714,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:54:41-07:00",
   "cost_usd": 0.0028,
   "latency_s": 21.1365,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:54:45-07:00",
   "cost_usd": 0.0824,
   "latency_s": 198.7333,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:13:0",
   "case_id": "13",
   "kind": "model_run",
   "title": "EU AI Act for AI sales platform",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:54:50-07:00",
   "cost_usd": 0.0046,
   "latency_s": 40.9394,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/13.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.0,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:54:52-07:00",
   "cost_usd": 0.109,
   "latency_s": 241.9905,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:55:10-07:00",
   "cost_usd": 0.0036,
   "latency_s": 27.8962,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:55:14-07:00",
   "cost_usd": 0.1032,
   "latency_s": 229.8462,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:claude-haiku-5.5:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 1.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 3.0,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:55:20-07:00",
   "cost_usd": 0.0076,
   "latency_s": 62.032,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/anthropic_claude-haiku-5.5/01.md",
   "notes": "The clean amendment renames the source-defined term \"Safe\" as \"Existing Safe\" throughout its operative provisions, contrary to the defined-term preservation requirement.",
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.0,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:55:47-07:00",
   "cost_usd": 0.1274,
   "latency_s": 282.9547,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 2.857,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:55:57-07:00",
   "cost_usd": 0.1342,
   "latency_s": 298.1454,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/01.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:grok-4.7:13:0",
   "case_id": "13",
   "kind": "model_run",
   "title": "EU AI Act for AI sales platform",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:56:07-07:00",
   "cost_usd": 0.1213,
   "latency_s": 282.8966,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/x-ai_grok-4.7/13.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:56:59-07:00",
   "cost_usd": 0.0242,
   "latency_s": 53.7,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "bakeoff:v2.6:gpt-6.1-terra:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": "pack v2.6",
   "run": "bakeoff",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.0,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T05:57:00-07:00",
   "cost_usd": 0.028,
   "latency_s": 61.4,
   "link": "evals/runs/2026-10-31-measured/outputs/v2.6/openai_gpt-6.1-terra/01.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:gpt-6-astra:CG3:0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:01:08-07:00",
   "cost_usd": 0.1351,
   "latency_s": 62.1776,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/openai_gpt-6-astra/CG3.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:gpt-6-astra:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 2.0,
   "polish": 1.0,
   "e2e_wow": 1.0,
   "task_score": 1.571,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:01:35-07:00",
   "cost_usd": 0.219,
   "latency_s": 88.4778,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/openai_gpt-6-astra/CG1.md",
   "notes": "Replaces the user's defined terms \"SAFEs\" and \"MFN Provision\" with \"Original SAFEs\" and \"MFN Rights\" in the replacement signature draft accompanying the redline, and drops the defined term \"Subsequent",
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:gpt-6-astra:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 3.0,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:01:36-07:00",
   "cost_usd": 0.1889,
   "latency_s": 89.067,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/openai_gpt-6-astra/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:gpt-6-astra:FG-L1:0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:01:43-07:00",
   "cost_usd": 0.213,
   "latency_s": 96.0725,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/openai_gpt-6-astra/FG-L1.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:gpt-6-astra:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 1.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:01:45-07:00",
   "cost_usd": 0.2302,
   "latency_s": 97.978,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/openai_gpt-6-astra/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:gpt-6-astra:01:0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 1.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 1.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 3.0,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:02-07:00",
   "cost_usd": 0.3766,
   "latency_s": 174.8776,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/openai_gpt-6-astra/01.md",
   "notes": "Renames the source-defined term \"Safe\" to \"Original SAFE\" in the operative standalone amendment and omits \"Subsequent Convertible Securities\", contrary to the defined-term preservation requirement.",
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:CG2:0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 1.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:30-07:00",
   "cost_usd": 0.0283,
   "latency_s": 44.8771,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/CG2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:CG4:0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:42-07:00",
   "cost_usd": 0.0284,
   "latency_s": 55.5208,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/CG4.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:CG1:0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 2.0,
   "polish": 1.0,
   "e2e_wow": 1.0,
   "task_score": 1.571,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:45-07:00",
   "cost_usd": 0.0375,
   "latency_s": 60.7397,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/CG1.md",
   "notes": "Replaces the user's document rather than preserving its defined terms in the requested redline: renames 'SAFEs' to 'Existing SAFEs' and drops 'Waiver', 'MFN Provision' and 'Subsequent Convertible Secu",
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:FG-L2:0",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 1.0,
   "grounding": 0.0,
   "specificity": 1.0,
   "actionability": 0.0,
   "polish": 4.0,
   "e2e_wow": 0.0,
   "task_score": 0.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:55-07:00",
   "cost_usd": 0.0037,
   "latency_s": 4.6847,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/FG-L2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:R2:0",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 1.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:17:31-07:00",
   "cost_usd": 0.0273,
   "latency_s": 52.0039,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/R2.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:05:0",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 2.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:17:50-07:00",
   "cost_usd": 0.0206,
   "latency_s": 46.0408,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/05.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:03:0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:18-07:00",
   "cost_usd": 0.044,
   "latency_s": 82.3164,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/03.md",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:18:0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 5.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:33-07:00",
   "cost_usd": 0.0184,
   "latency_s": 34.1245,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/18.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:11:0",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 1.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:19:03-07:00",
   "cost_usd": 0.0447,
   "latency_s": 82.5921,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/11.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:22:0",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 1.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 1.0,
   "e2e_wow": 2.0,
   "task_score": 2.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:19:37-07:00",
   "cost_usd": 0.0326,
   "latency_s": 69.0591,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/22.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "baseline-vs-pack:baseline-quality-bar:pascal-1:23:0",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "production QUALITY_BAR",
   "run": "baseline-vs-pack",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 2.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:20:36-07:00",
   "cost_usd": 0.0552,
   "latency_s": 112.3884,
   "link": "evals/runs/2026-10-31-measured/outputs/baseline-quality-bar/pascal-1/23.md",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG1-0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 2.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG1-1",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 2.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.0,
   "weight": 4,
   "hard_fail": true,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG3-0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG3-1",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-R1-0",
   "case_id": "R1",
   "kind": "model_run",
   "title": "Regression: Dubai NDA stall",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:23-07:00",
   "cost_usd": 0.0309,
   "latency_s": 58.0,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-R1-1",
   "case_id": "R1",
   "kind": "model_run",
   "title": "Regression: Dubai NDA stall",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:24-07:00",
   "cost_usd": 0.0317,
   "latency_s": 59.0,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-R3-0",
   "case_id": "R3",
   "kind": "model_run",
   "title": "Regression: intake overload",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:45-07:00",
   "cost_usd": 0.0113,
   "latency_s": 22.6,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-R3-1",
   "case_id": "R3",
   "kind": "model_run",
   "title": "Regression: intake overload",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:45-07:00",
   "cost_usd": 0.011,
   "latency_s": 23.0,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-FG-L1-0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-FG-L1-1",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-FG-L3-0",
   "case_id": "FG-L3",
   "kind": "model_run",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:37-07:00",
   "cost_usd": 0.0242,
   "latency_s": 14.5,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-FG-L3-1",
   "case_id": "FG-L3",
   "kind": "model_run",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:35-07:00",
   "cost_usd": 0.0082,
   "latency_s": 13.5,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-01-0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-01-1",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-06-0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-06-1",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:02:14-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-08-0",
   "case_id": "08",
   "kind": "model_run",
   "title": "SAFE conversion pro forma",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 2.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:30-07:00",
   "cost_usd": 0.0413,
   "latency_s": 65.9,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-08-1",
   "case_id": "08",
   "kind": "model_run",
   "title": "SAFE conversion pro forma",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": 0.0545,
   "latency_s": 90.1,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG2-0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG2-1",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG4-0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-CG4-1",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-R2-0",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:06-07:00",
   "cost_usd": 0.0336,
   "latency_s": 59.4,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-R2-1",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:15-07:00",
   "cost_usd": 0.0367,
   "latency_s": 67.1,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-FG-L2-0",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 0.0,
   "polish": 4.0,
   "e2e_wow": 1.0,
   "task_score": 1.714,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:04:12-07:00",
   "cost_usd": 0.0032,
   "latency_s": 5.6,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-FG-L2-1",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 0.0,
   "grounding": 0.0,
   "specificity": 1.0,
   "actionability": 0.0,
   "polish": 4.0,
   "e2e_wow": 0.0,
   "task_score": 0.714,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:04:09-07:00",
   "cost_usd": 0.0036,
   "latency_s": 4.9,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-03-0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "optimizer:v1-pascal-1-03-1",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:03:57-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "optimizer:v1-pascal-1-05-0",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:12-07:00",
   "cost_usd": 0.0308,
   "latency_s": 65.9,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-05-1",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:06-07:00",
   "cost_usd": 0.0291,
   "latency_s": 59.9,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-11-0",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:29-07:00",
   "cost_usd": 0.0453,
   "latency_s": 79.1,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-11-1",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:42-07:00",
   "cost_usd": 0.0549,
   "latency_s": 93.4,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-18-0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 5.0,
   "task_score": 4.714,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:04:09-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-18-1",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 5.0,
   "task_score": 4.714,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:04:09-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-22-0",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:30-07:00",
   "cost_usd": 0.0372,
   "latency_s": 71.7,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-22-1",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:05:32-07:00",
   "cost_usd": 0.035,
   "latency_s": 69.1,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-23-0",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.857,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:07:11-07:00",
   "cost_usd": 0.0568,
   "latency_s": 112.0,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v1-pascal-1-23-1",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v1",
   "run": "optimizer-v1",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:07:03-07:00",
   "cost_usd": 0.0563,
   "latency_s": 108.0,
   "link": "evals/optimizer/outputs/v1/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG1-0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.857,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:38-07:00",
   "cost_usd": 0.0321,
   "latency_s": 55.9,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG1-1",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:31-07:00",
   "cost_usd": 0.0273,
   "latency_s": 44.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG3-0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:43-07:00",
   "cost_usd": 0.0471,
   "latency_s": 59.5,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG3-1",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.571,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:38-07:00",
   "cost_usd": 0.0322,
   "latency_s": 55.0,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-R1-0",
   "case_id": "R1",
   "kind": "model_run",
   "title": "Regression: Dubai NDA stall",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:43-07:00",
   "cost_usd": 0.0326,
   "latency_s": 60.3,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-R1-1",
   "case_id": "R1",
   "kind": "model_run",
   "title": "Regression: Dubai NDA stall",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:45-07:00",
   "cost_usd": 0.0324,
   "latency_s": 62.3,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-R3-0",
   "case_id": "R3",
   "kind": "model_run",
   "title": "Regression: intake overload",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:02-07:00",
   "cost_usd": 0.0092,
   "latency_s": 20.2,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-R3-1",
   "case_id": "R3",
   "kind": "model_run",
   "title": "Regression: intake overload",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:01-07:00",
   "cost_usd": 0.009,
   "latency_s": 18.3,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-FG-L1-0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:09:11-07:00",
   "cost_usd": 0.0351,
   "latency_s": 61.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-FG-L1-1",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:09:05-07:00",
   "cost_usd": 0.0304,
   "latency_s": 54.2,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-FG-L3-0",
   "case_id": "FG-L3",
   "kind": "model_run",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:08:55-07:00",
   "cost_usd": 0.0106,
   "latency_s": 18.7,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-FG-L3-1",
   "case_id": "FG-L3",
   "kind": "model_run",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:09:00-07:00",
   "cost_usd": 0.0102,
   "latency_s": 17.8,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-01-0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:10:15-07:00",
   "cost_usd": 0.0501,
   "latency_s": 85.1,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-01-1",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.0,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:10:28-07:00",
   "cost_usd": 0.0563,
   "latency_s": 95.1,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-06-0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:09:32-07:00",
   "cost_usd": 0.0199,
   "latency_s": 39.9,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-06-1",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:09:41-07:00",
   "cost_usd": 0.0215,
   "latency_s": 47.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-08-0",
   "case_id": "08",
   "kind": "model_run",
   "title": "SAFE conversion pro forma",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.714,
   "weight": 2.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:10:02-07:00",
   "cost_usd": 0.035,
   "latency_s": 53.2,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-08-1",
   "case_id": "08",
   "kind": "model_run",
   "title": "SAFE conversion pro forma",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "train",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 2.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:10:16-07:00",
   "cost_usd": 0.0387,
   "latency_s": 64.1,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG2-0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:11:42-07:00",
   "cost_usd": 0.0359,
   "latency_s": 62.7,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG2-1",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.714,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:11:38-07:00",
   "cost_usd": 0.0391,
   "latency_s": 60.9,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG4-0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:11:40-07:00",
   "cost_usd": 0.0326,
   "latency_s": 59.5,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-CG4-1",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:11:25-07:00",
   "cost_usd": 0.0267,
   "latency_s": 50.2,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-R2-0",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:11:27-07:00",
   "cost_usd": 0.0272,
   "latency_s": 49.4,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-R2-1",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:11:33-07:00",
   "cost_usd": 0.0309,
   "latency_s": 54.9,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-FG-L2-0",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 0.0,
   "specificity": 1.0,
   "actionability": 0.0,
   "polish": 4.0,
   "e2e_wow": 0.0,
   "task_score": 1.0,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:10:41-07:00",
   "cost_usd": 0.0034,
   "latency_s": 6.8,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-FG-L2-1",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 4.0,
   "e2e_wow": 1.0,
   "task_score": 1.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:10:43-07:00",
   "cost_usd": 0.0038,
   "latency_s": 6.7,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-03-0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:11-07:00",
   "cost_usd": 0.0481,
   "latency_s": 80.1,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "optimizer:v2-pascal-1-03-1",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:06-07:00",
   "cost_usd": 0.0475,
   "latency_s": 75.1,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "optimizer:v2-pascal-1-05-0",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:54-07:00",
   "cost_usd": 0.0339,
   "latency_s": 80.2,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-05-1",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:44-07:00",
   "cost_usd": 0.0328,
   "latency_s": 67.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-11-0",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:49-07:00",
   "cost_usd": 0.0391,
   "latency_s": 66.2,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-11-1",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:13:09-07:00",
   "cost_usd": 0.0444,
   "latency_s": 80.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-18-0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:31-07:00",
   "cost_usd": 0.0235,
   "latency_s": 41.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-18-1",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:12:45-07:00",
   "cost_usd": 0.0279,
   "latency_s": 51.9,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-22-0",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:13:37-07:00",
   "cost_usd": 0.0424,
   "latency_s": 80.8,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-22-1",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:13:48-07:00",
   "cost_usd": 0.0449,
   "latency_s": 88.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-23-0",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:14:45-07:00",
   "cost_usd": 0.0607,
   "latency_s": 124.6,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v2-pascal-1-23-1",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v2",
   "run": "optimizer-v2",
   "split": "heldout",
   "sample": 1,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.857,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:14:56-07:00",
   "cost_usd": 0.0593,
   "latency_s": 121.5,
   "link": "evals/optimizer/outputs/v2/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-CG1-0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:22-07:00",
   "cost_usd": 0.0223,
   "latency_s": 37.9,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-CG3-0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.429,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:29-07:00",
   "cost_usd": 0.0243,
   "latency_s": 44.0,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-R1-0",
   "case_id": "R1",
   "kind": "model_run",
   "title": "Regression: Dubai NDA stall",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:57-07:00",
   "cost_usd": 0.0519,
   "latency_s": 71.1,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-R3-0",
   "case_id": "R3",
   "kind": "model_run",
   "title": "Regression: intake overload",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.571,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:07-07:00",
   "cost_usd": 0.0115,
   "latency_s": 23.0,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-FG-L1-0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:31-07:00",
   "cost_usd": 0.0251,
   "latency_s": 46.2,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-FG-L3-0",
   "case_id": "FG-L3",
   "kind": "model_run",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:15:54-07:00",
   "cost_usd": 0.0081,
   "latency_s": 14.1,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-01-0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:17:25-07:00",
   "cost_usd": 0.0605,
   "latency_s": 100.2,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-06-0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:16:30-07:00",
   "cost_usd": 0.0209,
   "latency_s": 45.6,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-08-0",
   "case_id": "08",
   "kind": "model_run",
   "title": "SAFE conversion pro forma",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.857,
   "weight": 2.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:17:12-07:00",
   "cost_usd": 0.0374,
   "latency_s": 66.1,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-CG2-0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:32-07:00",
   "cost_usd": 0.0352,
   "latency_s": 59.9,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-CG4-0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:28-07:00",
   "cost_usd": 0.03,
   "latency_s": 54.6,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-R2-0",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:25-07:00",
   "cost_usd": 0.029,
   "latency_s": 50.2,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-FG-L2-0",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 4.0,
   "e2e_wow": 1.0,
   "task_score": 2.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:17:40-07:00",
   "cost_usd": 0.0029,
   "latency_s": 4.7,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-03-0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:19:11-07:00",
   "cost_usd": 0.0573,
   "latency_s": 96.4,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "optimizer:v3-pascal-1-05-0",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:42-07:00",
   "cost_usd": 0.0302,
   "latency_s": 66.7,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-11-0",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:42-07:00",
   "cost_usd": 0.0382,
   "latency_s": 68.1,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-18-0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:16-07:00",
   "cost_usd": 0.023,
   "latency_s": 41.6,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-22-0",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:18:58-07:00",
   "cost_usd": 0.0321,
   "latency_s": 66.6,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v3-pascal-1-23-0",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v3",
   "run": "optimizer-v3",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:20:23-07:00",
   "cost_usd": 0.057,
   "latency_s": 115.9,
   "link": "evals/optimizer/outputs/v3/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-CG1-0",
   "case_id": "CG1",
   "kind": "model_run",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.571,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:21:50-07:00",
   "cost_usd": 0.0266,
   "latency_s": 40.5,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-CG3-0",
   "case_id": "CG3",
   "kind": "model_run",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.429,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:21:58-07:00",
   "cost_usd": 0.0302,
   "latency_s": 48.8,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-R1-0",
   "case_id": "R1",
   "kind": "model_run",
   "title": "Regression: Dubai NDA stall",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.286,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:22:10-07:00",
   "cost_usd": 0.0327,
   "latency_s": 60.9,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-R3-0",
   "case_id": "R3",
   "kind": "model_run",
   "title": "Regression: intake overload",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 5.0,
   "task_score": 4.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:21:26-07:00",
   "cost_usd": 0.0256,
   "latency_s": 20.7,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-FG-L1-0",
   "case_id": "FG-L1",
   "kind": "model_run",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:22:02-07:00",
   "cost_usd": 0.0291,
   "latency_s": 50.3,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-FG-L3-0",
   "case_id": "FG-L3",
   "kind": "model_run",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.857,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:21:28-07:00",
   "cost_usd": 0.0113,
   "latency_s": 19.0,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-01-0",
   "case_id": "01",
   "kind": "model_run",
   "title": "MFN side letter waiver + cover email",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 3.0,
   "task_score": 3.0,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:22:28-07:00",
   "cost_usd": 0.0479,
   "latency_s": 75.5,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-06-0",
   "case_id": "06",
   "kind": "model_run",
   "title": "New Associate from free text",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:21:56-07:00",
   "cost_usd": 0.0227,
   "latency_s": 47.0,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-08-0",
   "case_id": "08",
   "kind": "model_run",
   "title": "SAFE conversion pro forma",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "train",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:22:43-07:00",
   "cost_usd": 0.04,
   "latency_s": 66.2,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-CG2-0",
   "case_id": "CG2",
   "kind": "model_run",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 4,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:23:50-07:00",
   "cost_usd": 0.0369,
   "latency_s": 56.8,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-CG4-0",
   "case_id": "CG4",
   "kind": "model_run",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 3,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:23:43-07:00",
   "cost_usd": 0.0304,
   "latency_s": 51.5,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-R2-0",
   "case_id": "R2",
   "kind": "model_run",
   "title": "Regression: California small claims self-help",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.429,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:23:37-07:00",
   "cost_usd": 0.0274,
   "latency_s": 44.7,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-FG-L2-0",
   "case_id": "FG-L2",
   "kind": "model_run",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 4.0,
   "e2e_wow": 1.0,
   "task_score": 2.0,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:22:58-07:00",
   "cost_usd": 0.0032,
   "latency_s": 5.3,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-03-0",
   "case_id": "03",
   "kind": "model_run",
   "title": "DPA redline (customer-derived, Art. 28)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.286,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:24:11-07:00",
   "cost_usd": 0.0485,
   "latency_s": 77.0,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "optimizer:v4-pascal-1-05-0",
   "case_id": "05",
   "kind": "model_run",
   "title": "California custody move-away",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:24:01-07:00",
   "cost_usd": 0.0365,
   "latency_s": 69.6,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-11-0",
   "case_id": "11",
   "kind": "model_run",
   "title": "Enterprise MSA counter-redline (NY)",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.143,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:24:05-07:00",
   "cost_usd": 0.0427,
   "latency_s": 72.5,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-18-0",
   "case_id": "18",
   "kind": "model_run",
   "title": "California security deposit",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 5.0,
   "task_score": 4.857,
   "weight": 2.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:23:38-07:00",
   "cost_usd": 0.0262,
   "latency_s": 45.1,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-22-0",
   "case_id": "22",
   "kind": "model_run",
   "title": "Series A term sheet counter",
   "source": "authored, fictional parties",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.429,
   "weight": 3.0,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:24:14-07:00",
   "cost_usd": 0.0334,
   "latency_s": 64.4,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "optimizer:v4-pascal-1-23-0",
   "case_id": "23",
   "kind": "model_run",
   "title": "EU shop privacy notice + cookies",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": "optimizer v4",
   "run": "optimizer-v4",
   "split": "heldout",
   "sample": 0,
   "scale": "0-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 4.143,
   "weight": 1.5,
   "hard_fail": false,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": "pascal-1",
   "run_date": "2026-10-31T06:25:55-07:00",
   "cost_usd": 0.0671,
   "latency_s": 128.2,
   "link": "evals/optimizer/outputs/v4/",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:01-mfn-side-letter-edit",
   "case_id": "01-mfn-side-letter-edit",
   "kind": "task_score",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 2.14,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Real analog run 3cee9a8e ([name] MFN, 11:41 UTC): dilution right per SAFE (9.64%, 14.55%) but combined figure compounds them (22.75%) which is wrong for post-money SAFEs (~24.2%); MFN answer hedged; Names DGCL generally; no pinpoints beyond s.141 style mentions, no links, no Sources list; no retrieval in pipeline; Uses names and figures; but general chat uses the commercial checklist (selectDra",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:02-terse-term-sheet-and-safe",
   "case_id": null,
   "kind": "run_summary",
   "title": "02-terse-term-sheet-and-safe",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 2.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 2.14,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "[name] SAFE run 20a8acb7 shows sound US financing law; terse \"2 and 5m post\" parsing untested; no intent router in code; [name] run cites Rule 506(b), Form D 15 days, DGCL 141(f), 152-153 correctly; no links; Terse ask will trigger an intake card (2 to 5 questions) before any draft; Likely produces advice plus plan, not a full term sheet in one pass; Long advisory prose; no em dashes; Send intent not ",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:03-dpa-redline-customer-derived",
   "case_id": "03-dpa-redline-customer-derived",
   "kind": "task_score",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Privacy checklist covers Art 28 items, transfers, breach; SLA plus EU residency run 10:45 UTC is competent; Art 28 named, no (3)(a) to (h) mapping enforced, no SCC decision number or links; customer-derived app playbook row only says \"Art. 28 terms\"; Redline format quotes source text with deletions and insertions; Redline plus fallback plus negotiation email bundle; Clean but verbose; Good bundle; customer-derived pl",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:04-sponsorship-customer-derived",
   "case_id": "04-sponsorship-customer-derived",
   "kind": "task_score",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.57,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Falls to commercial checklist; no sports or sponsorship points (category exclusivity, make-goods, ambush); No sector anchors; risk of generic citations; Will quote clauses but miss rights-holder specific positions; Redline plus fallback bundle exists; Clean but verbose; customer-derived app has no sponsorship playbook either; weighted, e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:05-family-law-california-custody",
   "case_id": null,
   "kind": "run_summary",
   "title": "05-family-law-california-custody",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.71,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Measured analog (E&W divorce 0b212c3d, judge 7.4): correct MCA 1973 s.25, CMS, MPS, LSPO; Real statute named, no links, no Sources list; Uses facts but no intent router: non-contract matters (divorce; Action plan present, buried in overlapping documents; Judge concision 6.3; stale cross references between extras; Simulated counsel emails misidentified the client (judge notes); misrouting; weighted",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:06-new-associate-from-matter",
   "case_id": null,
   "kind": "run_summary",
   "title": "06-new-associate-from-matter",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 1.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 1.29,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "No create-Associate capability: bots.static.ts is a fixed list; no route or server function creates a bot; Request would be treated as a contract review of a missing document; No path to a working Associate; Copy itself would be clean; Biggest missing wow moment; weighted, e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:07-delaware-incorporation",
   "case_id": null,
   "kind": "run_summary",
   "title": "07-delaware-incorporation",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Corporate checklist plus DGCL and 83(b) lines in QUALITY_BAR; Real cites likely, no links; Good; Checklist with steps; Verbose; Misroute risk to Contract Review (seen for Dubai incorporation); weighted, e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:analog:customer-derived-mfn-demo 61e41a94",
   "case_id": null,
   "kind": "run_summary",
   "title": "analog:customer-derived-mfn-demo 61e41a94",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 2.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 4.0,
   "e2e_wow": 1.0,
   "task_score": 1.57,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Answered bank data breach liability instead of MFN. Hard fail.",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:analog:customer-derived-mfn-real 3cee9a8e",
   "case_id": null,
   "kind": "run_summary",
   "title": "analog:customer-derived-mfn-real 3cee9a8e",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 1.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 1.86,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Initial advice plus intake only; compounding dilution error; stalled needs_input.",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:analog:customer-derived-safe-real 20a8acb7",
   "case_id": null,
   "kind": "run_summary",
   "title": "analog:customer-derived-safe-real 20a8acb7",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Sound Reg D, Form D, DGCL; CA notice fee stated flat $300 needs verification; long.",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-baseline:analog:ew-divorce-real 0b212c3d",
   "case_id": null,
   "kind": "run_summary",
   "title": "analog:ew-divorce-real 0b212c3d",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 3.0,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Correct law; misrouted to Contract Review; overlapping docs.",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:01-mfn-side-letter-edit",
   "case_id": "01-mfn-side-letter-edit",
   "kind": "task_score",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.86,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:02-terse-term-sheet-and-safe",
   "case_id": null,
   "kind": "run_summary",
   "title": "02-terse-term-sheet-and-safe",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.86,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:03-dpa-redline-customer-derived",
   "case_id": "03-dpa-redline-customer-derived",
   "kind": "task_score",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.86,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:04-sponsorship-customer-derived",
   "case_id": "04-sponsorship-customer-derived",
   "kind": "task_score",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:05-family-law-california-custody",
   "case_id": null,
   "kind": "run_summary",
   "title": "05-family-law-california-custody",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.86,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:06-new-associate-from-matter",
   "case_id": null,
   "kind": "run_summary",
   "title": "06-new-associate-from-matter",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:2026-10-31-after-projected:07-delaware-incorporation",
   "case_id": null,
   "kind": "run_summary",
   "title": "07-delaware-incorporation",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "2026-10-31-after-projected",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.57,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (projected, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written; assumes P1 to P3, M1, X1 land as written",
   "segment": null
  },
  {
   "row_id": "scores:GC-C0-production:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C0-production",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.56,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "2",
   "basis": "measured",
   "label": "MEASURED (production, from 31 Oct DB runs read-only)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "12:05 UTC runs: Dubai NDA went to nda, LA small claims to advice; earlier 10 of 60 misrouted to Contract Review; Small claims understood, but tenant called \"the founder\" (e86043d0); 4 to 5 questions per run; task blocks in needs_input before drafting; Law correct: Civ. Code 1950.5 21 days, 2x bad faith, $12,500 limit; Hedge walls, \"Founder action and exposure\" for a tenant, long overlapping docs; ",
   "segment": null
  },
  {
   "row_id": "scores:GC-C1-v2.0:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C1-v2.0",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.22,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "3",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.0, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Only 3 router lines; no examples, no tie-breaks; No audience detection (individual vs founder); 3 max and non-blocking, but no rule for which 3; Same law guidance as production; Good voice rules; No concrete fix; prompt cannot change pipeline timing; Partial list; prompt cannot strip server text; Sources named but no format or confidence; Only small claims covered",
   "segment": null
  },
  {
   "row_id": "scores:GC-C2-v2.1:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C2-v2.1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.89,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "4",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.1, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Full mapping table with examples and tie-breaks; Audience rule fixes founder misuse; Selection rule: only questions that change the answer; Matter rules added; Voice rules plus length caps; Code fixes listed for stream claim and lawyer loop; Full leak list plus server strip code fix; Sources block with confidence and verify labels; Self-draft rule and lawyer-only-for-judgment rule",
   "segment": null
  },
  {
   "row_id": "scores:GC-C3-v2.2:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C3-v2.2",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.22,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.2, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Same table plus associate and send flags; Terse money parsing, send intent, edit-own-doc, associate requests; Zero questions when answer does not change; chips; beside the draft; Same; Same; Same code fixes; prompt adds 90s first-answer target; Same; Confidence calibration rule added; Lawyer step after deliverable, never for small claims or simple asks",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-baseline:CG1",
   "case_id": "CG1",
   "kind": "task_score",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 2.0,
   "specificity": 4.0,
   "actionability": 1.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 2.29,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (real run c6433845)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Caught majority-binds-all risk, but stopped in needs_input asking for cap table and all SAFE forms; no redline, no clean version, no email, no signing; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-baseline:CG2",
   "case_id": "CG2",
   "kind": "task_score",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 2.0,
   "specificity": 5.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 3.14,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (real run 2dd5a0f5)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Caught s.2 threshold and managing partner consent; scoped waiver with defined terms; no Sources; \"Founder summary\" and raw \"FDL sign off required:\"; no clean version, cover email or signing; sibling run 33e37484 blocked in intake; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-baseline:CG3",
   "case_id": "CG3",
   "kind": "task_score",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 2.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 2.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (no matching real run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "NDA bot redlines well in other runs; no signing flow in main project; lawyer loop adds minutes; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-baseline:CG4",
   "case_id": "CG4",
   "kind": "task_score",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 2.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (no matching real run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Corporate checklist has DGCL lines; no signing flow; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-baseline:COLD overall (CG1 4, CG2 4, CG3 3, CG4 3)",
   "case_id": null,
   "kind": "run_summary",
   "title": "COLD overall (CG1 4, CG2 4, CG3 3, CG4 3)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.65,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "see rows",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-prompt-only:CG1",
   "case_id": "CG1",
   "kind": "task_score",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 2.0,
   "polish": 4.0,
   "e2e_wow": 2.0,
   "task_score": 3.0,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Prompt says never block, but the intake gate is code (fix 9), so the run can still stop; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-prompt-only:CG2",
   "case_id": "CG2",
   "kind": "task_score",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.86,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Clean version, pushback and cover email asked for by prompt; signing still missing; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-prompt-only:CG3",
   "case_id": "CG3",
   "kind": "task_score",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 2.0,
   "task_score": 3.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Same; no envelope; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-prompt-only:CG4",
   "case_id": "CG4",
   "kind": "task_score",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 2.0,
   "task_score": 3.43,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Same; no envelope; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-prompt-only:COLD overall (CG1 4, CG2 4, CG3 3, CG4 3)",
   "case_id": null,
   "kind": "run_summary",
   "title": "COLD overall (CG1 4, CG2 4, CG3 3, CG4 3)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.4,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "see rows",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-plus-code-fixes:CG1",
   "case_id": "CG1",
   "kind": "task_score",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static, assumes code fixes 0, 9 to 12 and Founder acceptance path signing)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "End to end in chat with simulated envelope; e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-plus-code-fixes:CG2",
   "case_id": "CG2",
   "kind": "task_score",
   "title": "Cold: SAFE side letter MFN amendment",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.29,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (same assumptions)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-plus-code-fixes:CG3",
   "case_id": "CG3",
   "kind": "task_score",
   "title": "Cold: NDA change and send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (same assumptions)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-plus-code-fixes:CG4",
   "case_id": "CG4",
   "kind": "task_score",
   "title": "Cold: board consent for a SAFE round",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (same assumptions)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4-plus-code-fixes:COLD overall (CG1 4, CG2 4, CG3 3, CG4 3)",
   "case_id": null,
   "kind": "run_summary",
   "title": "COLD overall (CG1 4, CG2 4, CG3 3, CG4 3)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.12,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "see rows",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C5-v2.4:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C5-v2.4",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.33,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.4, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "v2.3 known traps lift quality; speed and leaks still need code fixes 0, 1, 6",
   "segment": null
  },
  {
   "row_id": "scores:GC-C4-v2.3:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C4-v2.3",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.33,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.3, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Known traps list lifts quality from 4 to 5; logged after C5 by mistake, same values",
   "segment": null
  },
  {
   "row_id": "scores:GC-C6-v2.5:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C6-v2.5",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.33,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.5, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Same scores as C5; removes the QUALITY_BAR conflict that would undo leak and tone gains",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-0531:CG1",
   "case_id": "CG1",
   "kind": "task_score",
   "title": "Cold: MFN waiver upload, change, send for signature (simulated)",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-0531",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.86,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (real runs df59fc57, dba1a0ff, follow-up 0cab8f64; read-only DB, 5:22 to 5:25 AM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Turn 1 now redlines without demanding the cap table: 2026 scope and 31 Dec 2026 sunset in both drafts, done in about 4 min. Still: FDL sign off / no paid work leaks in 5 of 8 docs, Commercial drafter and a UK SaaS lawyer for a Delaware SAFE, follow-up edit (signing per holder, drop majority-binds-all) blocked in needs_input with 6 facts missing. e2e_wow x2",
   "segment": null
  },
  {
   "row_id": "scores:GC-COLD-0531:cold overall",
   "case_id": null,
   "kind": "run_summary",
   "title": "cold overall",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-COLD-0531",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.82,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "MIXED (CG1 measured 0531, CG2 measured, CG3 and CG4 estimates)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "was 2.65; weights CG1 4, CG2 4, CG3 3, CG4 3",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-baseline:FG-L1",
   "case_id": "FG-L1",
   "kind": "session_score",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.86,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 1 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Correct GDPR and UCTA, 12/12 cites correct, but DORA only when asked by name",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-baseline:FG-L2",
   "case_id": "FG-L2",
   "kind": "session_score",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.57,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 1 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "5 intake questions plus upload before any answer",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-baseline:FG-L3",
   "case_id": "FG-L3",
   "kind": "session_score",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 1.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 1.86,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 1 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Declined lawyer quotes on its own, task Done with FDL sign off outstanding. DB: 264 fee decisions in 4 h, 0 asked the user",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-baseline:FG-L4",
   "case_id": "FG-L4",
   "kind": "session_score",
   "title": "First view under 10 seconds",
   "source": "focus group, anonymised real inputs",
   "use_case": "Latency",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 2.29,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 1 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "55 to 80 s perceived first view; server first view p50 6.8 s, p90 10.2 s; first question lost to profile modal",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6-prompt-only:FG-L1",
   "case_id": "FG-L1",
   "kind": "task_score",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.6, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Trigger table 1d",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6-prompt-only:FG-L2",
   "case_id": "FG-L2",
   "kind": "task_score",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.6, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Answer first 1e",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6-prompt-only:FG-L3",
   "case_id": "FG-L3",
   "kind": "task_score",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.57,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.6, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Prompt asks for consent but orchestrator still auto-declines; needs fix 14",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6-prompt-only:FG-L4",
   "case_id": "FG-L4",
   "kind": "task_score",
   "title": "First view under 10 seconds",
   "source": "focus group, anonymised real inputs",
   "use_case": "Latency",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 2.29,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.6, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Prompt cannot fix modal; needs fix 16",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6-plus-code-fixes:FG-L3",
   "case_id": "FG-L3",
   "kind": "task_score",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review, code fixes not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Fixes 14 and 15",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6-plus-code-fixes:FG-L4",
   "case_id": "FG-L4",
   "kind": "task_score",
   "title": "First view under 10 seconds",
   "source": "focus group, anonymised real inputs",
   "use_case": "Latency",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.71,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review, code fixes not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Fix 16; speed not measured",
   "segment": null
  },
  {
   "row_id": "scores:GC-C7-v2.6:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C7-v2.6",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.44,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.6, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Answer first; Fixed skeleton for consistency; Regulated counterparty table; was 4.33",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F1",
   "case_id": "FG-F1",
   "kind": "session_score",
   "title": "Series A MFN waiver: edit my docx, redline, sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.71,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Session 1: docx not used, follow-up new task, raw markdown. Session 8 one-shot was much better (redline in 51 s)",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F2",
   "case_id": "FG-F2",
   "kind": "session_score",
   "title": "First-time founder: SAFE dilution on a phone",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 1.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 2.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Refused the dilution number on turn 3, re-asked raise",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F3",
   "case_id": "FG-F3",
   "kind": "session_score",
   "title": "Texas unpaid invoice: demand letter now",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 1.0,
   "e2e_wow": 1.0,
   "task_score": 1.86,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "No ch. 38 or interest, asked invoice year, letter lost on turn 3",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F4",
   "case_id": "FG-F4",
   "kind": "session_score",
   "title": "Solo consultant: NDA top risks, sign today",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 2.0,
   "polish": 1.0,
   "e2e_wow": 2.0,
   "task_score": 2.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Strong first view, final summary asked for the attached NDA",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F5",
   "case_id": "FG-F5",
   "kind": "session_score",
   "title": "Divorce in England, two children, joint home",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 1.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 1.0,
   "e2e_wow": 1.0,
   "task_score": 1.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Lost Manchester, generic answer, QA 8.2/10",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F6",
   "case_id": "FG-F6",
   "kind": "session_score",
   "title": "Employee in California, HR meeting tomorrow",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 1.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 2.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Accurate substance, no cites, no UI Code 1089, sheet blocked composer",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F7",
   "case_id": "FG-F7",
   "kind": "session_score",
   "title": "Angel wants a board seat on a SAFE",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Right verdict, five business days and $1,050 cap on a quick question",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-L5",
   "case_id": "FG-L5",
   "kind": "session_score",
   "title": "Governing law in a data transfer context",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract redline",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 1.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 1.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived: California law vs UK bank, Sales can agree, 98%",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-L6",
   "case_id": "FG-L6",
   "kind": "session_score",
   "title": "Processor breach notice for a bank: 24 hours, not 72",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived: 72 hour processor target for a bank",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-L7",
   "case_id": "FG-L7",
   "kind": "session_score",
   "title": "Critical or important function: DORA Art. 30 mapping",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived: no DORA Art. 30, PRA or EBA",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-L8",
   "case_id": "FG-L8",
   "kind": "session_score",
   "title": "After flagging Privacy Shield, name the current route",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Caught Privacy Shield, did not name the DPF",
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F1",
   "case_id": "FG-F1",
   "kind": "task_score",
   "title": "Series A MFN waiver: edit my docx, redline, sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.86,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F2",
   "case_id": "FG-F2",
   "kind": "task_score",
   "title": "First-time founder: SAFE dilution on a phone",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.71,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F3",
   "case_id": "FG-F3",
   "kind": "task_score",
   "title": "Texas unpaid invoice: demand letter now",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.57,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F4",
   "case_id": "FG-F4",
   "kind": "task_score",
   "title": "Solo consultant: NDA top risks, sign today",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.71,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F5",
   "case_id": "FG-F5",
   "kind": "task_score",
   "title": "Divorce in England, two children, joint home",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 3.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 2.0,
   "polish": 2.0,
   "e2e_wow": 2.0,
   "task_score": 2.43,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F6",
   "case_id": "FG-F6",
   "kind": "task_score",
   "title": "Employee in California, HR meeting tomorrow",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.71,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F7",
   "case_id": "FG-F7",
   "kind": "task_score",
   "title": "Angel wants a board seat on a SAFE",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 3.0,
   "task_score": 3.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-L5",
   "case_id": "FG-L5",
   "kind": "task_score",
   "title": "Governing law in a data transfer context",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract redline",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "FG-L5 to L8 also need the customer-derived message",
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-L6",
   "case_id": "FG-L6",
   "kind": "task_score",
   "title": "Processor breach notice for a bank: 24 hours, not 72",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "FG-L5 to L8 also need the customer-derived message",
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-L7",
   "case_id": "FG-L7",
   "kind": "task_score",
   "title": "Critical or important function: DORA Art. 30 mapping",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "FG-L5 to L8 also need the customer-derived message",
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-L8",
   "case_id": "FG-L8",
   "kind": "task_score",
   "title": "After flagging Privacy Shield, name the current route",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7 prompt changes, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "FG-L5 to L8 also need the customer-derived message",
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F1",
   "case_id": "FG-F1",
   "kind": "task_score",
   "title": "Series A MFN waiver: edit my docx, redline, sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F2",
   "case_id": "FG-F2",
   "kind": "task_score",
   "title": "First-time founder: SAFE dilution on a phone",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F3",
   "case_id": "FG-F3",
   "kind": "task_score",
   "title": "Texas unpaid invoice: demand letter now",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F4",
   "case_id": "FG-F4",
   "kind": "task_score",
   "title": "Solo consultant: NDA top risks, sign today",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F5",
   "case_id": "FG-F5",
   "kind": "task_score",
   "title": "Divorce in England, two children, joint home",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F6",
   "case_id": "FG-F6",
   "kind": "task_score",
   "title": "Employee in California, HR meeting tomorrow",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F7",
   "case_id": "FG-F7",
   "kind": "task_score",
   "title": "Angel wants a board seat on a SAFE",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (code fixes 18 to 27 queued, not built)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.56,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.7, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Persona switch and quick question mode; was 4.44",
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-F avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.26,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-baseline-0550:FG-L5-8 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L5-8 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-baseline-0550",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.39,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-F avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.06,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-prompt-only:FG-L5-8 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L5-8 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-prompt-only",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.04,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C8-v2.7-plus-code-fixes:FG-F avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C8-v2.7-plus-code-fixes",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.1,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S3-baseline:FG-L9",
   "case_id": "FG-L9",
   "kind": "session_score",
   "title": "Honest review label until a lawyer signs off",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S3-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 2.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 4.0,
   "e2e_wow": 1.0,
   "task_score": 2.14,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 3 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived: false reviewed by your Index lawyer\" label in about 9 s (hard fail)\"",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S3-baseline:FG-L10",
   "case_id": "FG-L10",
   "kind": "session_score",
   "title": "EIS and SEIS connected persons",
   "source": "focus group, anonymised real inputs",
   "use_case": "Tax advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S3-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 1.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 1.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 3 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived: pitched 30% EIS to associates, s.253 tag everywhere",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S3-baseline:FG-L11",
   "case_id": "FG-L11",
   "kind": "session_score",
   "title": "One banking breach notice position across apps",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S3-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 3.0,
   "specificity": 3.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.57,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 3 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived 72 h default vs Pascal-1 Demo DPA 24 h",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S3-baseline:FG-L12",
   "case_id": "FG-L12",
   "kind": "session_score",
   "title": "UK fundraising: financial promotion restriction and exemptions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S3-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 2.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 2.14,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group Session 3 observations, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived: no s.21, art. 48 or 50A",
   "segment": null
  },
  {
   "row_id": "scores:GC-C9-v2.8:FG-L9",
   "case_id": "FG-L9",
   "kind": "task_score",
   "title": "Honest review label until a lawyer signs off",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C9-v2.8",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 4.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (pack v2.8 plus queued Pascal-1 and customer-derived messages; SEGC1 not queued)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Pascal-1 only; customer-derived needs SEGC1",
   "segment": null
  },
  {
   "row_id": "scores:GC-C9-v2.8:FG-L10",
   "case_id": "FG-L10",
   "kind": "task_score",
   "title": "EIS and SEIS connected persons",
   "source": "focus group, anonymised real inputs",
   "use_case": "Tax advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C9-v2.8",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (pack v2.8 plus queued Pascal-1 and customer-derived messages; SEGC1 not queued)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Pascal-1 only; customer-derived needs SEGC1",
   "segment": null
  },
  {
   "row_id": "scores:GC-C9-v2.8:FG-L11",
   "case_id": "FG-L11",
   "kind": "task_score",
   "title": "One banking breach notice position across apps",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C9-v2.8",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.29,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (pack v2.8 plus queued Pascal-1 and customer-derived messages; SEGC1 not queued)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "customer-derived G plus Pascal-1 F",
   "segment": null
  },
  {
   "row_id": "scores:GC-C9-v2.8:FG-L12",
   "case_id": "FG-L12",
   "kind": "task_score",
   "title": "UK fundraising: financial promotion restriction and exemptions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C9-v2.8",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 5.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.29,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (pack v2.8 plus queued Pascal-1 and customer-derived messages; SEGC1 not queued)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Pascal-1 only; customer-derived needs SEGC1",
   "segment": null
  },
  {
   "row_id": "scores:GC-C9-v2.8:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C9-v2.8",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.56,
   "weight": null,
   "hard_fail": null,
   "metric": "self_draft",
   "metric_value": "5",
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.8, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Adds UK fundraising rules; no dimension moves on the regression set; unchanged from C8",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S3-baseline:FG-L9-12 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L9-12 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S3-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.1,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C9-v2.8:FG-L9-12 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L9-12 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C9-v2.8",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.22,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S4-baseline:FG-L13 controller before DPA",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L13 controller before DPA",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S4-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 1.86,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored on Session 4 customer-derived output: acc 1, grnd 2, spec 2, act 2, pol 4, wow 1)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "called independent controller sharing a missing DPA",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S4-baseline:FG-L14 penalty clause check",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L14 penalty clause check",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S4-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored on Session 4 customer-derived output: acc 2, grnd 3, spec 2, act 2, pol 4, wow 2)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "140% and 100% clauses never tested as penalties",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S4-baseline:FG-L13-14 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L13-14 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S4-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.14,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C10-v2.9:FG-L13 controller before DPA",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L13 controller before DPA",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C10-v2.9",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.9 and customer-derived message H, no model run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C10-v2.9:FG-L14 penalty clause check",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L14 penalty clause check",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C10-v2.9",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.29,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "depends on governing law being read correctly per clause",
   "segment": null
  },
  {
   "row_id": "scores:GC-C10-v2.9:FG-L13-14 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L13-14 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C10-v2.9",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.36,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C10-v2.9:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C10-v2.9",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.56,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "unchanged; new rules only fire on data sharing and fixed sum clauses",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S5-baseline:FG-L15 CA work for hire cite",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L15 CA work for hire cite",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S5-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.29,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored on Session 5 output: acc 3, grnd 2, spec 4, act 4, pol 4, wow 3)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "cited s.3351(c) instead of s.3351.5(c)",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S5-baseline:FG-L16 NSI threshold cite",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L16 NSI threshold cite",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S5-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.29,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored on Session 5 output: acc 4, grnd 2, spec 3, act 4, pol 4, wow 3)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "thresholds right but uncited",
   "segment": null
  },
  {
   "row_id": "scores:FG-L-S5-baseline:FG-L15-16 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L15-16 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-L-S5-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.29,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-L15-16 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-L15-16 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (static review of pack v2.10; held in next-batch.md, not queued)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:general-chat (R1-R4, 01, 02, 06)",
   "case_id": null,
   "kind": "run_summary",
   "title": "general-chat (R1-R4, 01, 02, 06)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.56,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "unchanged; grounding gains only on threshold answers",
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F8",
   "case_id": "FG-F8",
   "kind": "session_score",
   "title": "Series A founder: one-shot MFN edit, clean, redline, e-sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.43,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 4 4 4 3 3 3)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F8",
   "case_id": "FG-F8",
   "kind": "task_score",
   "title": "Series A founder: one-shot MFN edit, clean, redline, e-sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.29,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F9",
   "case_id": "FG-F9",
   "kind": "session_score",
   "title": "Employee 44: severance release, 7 days to sign (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 1 2 3 3 3 1)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F9",
   "case_id": "FG-F9",
   "kind": "task_score",
   "title": "Employee 44: severance release, 7 days to sign (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F10",
   "case_id": "FG-F10",
   "kind": "session_score",
   "title": "Texas SMB: bounced $3,200 check, write the letter",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 3 2 3 2 3 2)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F10",
   "case_id": "FG-F10",
   "kind": "task_score",
   "title": "Texas SMB: bounced $3,200 check, write the letter",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.29,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F11",
   "case_id": "FG-F11",
   "kind": "session_score",
   "title": "England divorce: no-fault, timing, court (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.71,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 4 3 4 4 3 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F11",
   "case_id": "FG-F11",
   "kind": "task_score",
   "title": "England divorce: no-fault, timing, court (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F12",
   "case_id": "FG-F12",
   "kind": "session_score",
   "title": "California consultant: SOW red flags (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.71,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 3 3 4 4 4 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F12",
   "case_id": "FG-F12",
   "kind": "task_score",
   "title": "California consultant: SOW red flags (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F13",
   "case_id": "FG-F13",
   "kind": "session_score",
   "title": "Series A founder: MFN waiver on phone, Continue path",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 1.57,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 2 2 2 1 2 1)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F13",
   "case_id": "FG-F13",
   "kind": "task_score",
   "title": "Series A founder: MFN waiver on phone, Continue path",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F14",
   "case_id": "FG-F14",
   "kind": "session_score",
   "title": "Texas SMB: 'write it now' with all facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 3 2 3 3 3 2)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F14",
   "case_id": "FG-F14",
   "kind": "task_score",
   "title": "Texas SMB: 'write it now' with all facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F15",
   "case_id": "FG-F15",
   "kind": "session_score",
   "title": "Founder: 'just give me the number' (SAFE %)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.57,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 5 3 5 3 3 3)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F15",
   "case_id": "FG-F15",
   "kind": "task_score",
   "title": "Founder: 'just give me the number' (SAFE %)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.57,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F16",
   "case_id": "FG-F16",
   "kind": "session_score",
   "title": "Consultant: NDA with 'no questions' (desktop)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 4 4 4 3 2 3)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F16",
   "case_id": "FG-F16",
   "kind": "task_score",
   "title": "Consultant: NDA with 'no questions' (desktop)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F17",
   "case_id": "FG-F17",
   "kind": "session_score",
   "title": "England: threat to move children to Glasgow (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.86,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 4 3 4 4 4 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F17",
   "case_id": "FG-F17",
   "kind": "task_score",
   "title": "England: threat to move children to Glasgow (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F18",
   "case_id": "FG-F18",
   "kind": "session_score",
   "title": "San Jose employee: final pay on next payday",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 5 3 4 4 4 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F18",
   "case_id": "FG-F18",
   "kind": "task_score",
   "title": "San Jose employee: final pay on next payday",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.57,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F19",
   "case_id": "FG-F19",
   "kind": "session_score",
   "title": "Series A founder: clean doc plus investor emails",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 1.43,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 2 2 1 1 2 1)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F19",
   "case_id": "FG-F19",
   "kind": "task_score",
   "title": "Series A founder: clean doc plus investor emails",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F20",
   "case_id": "FG-F20",
   "kind": "session_score",
   "title": "Texas SMB: letter, then 'email it'",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 3 2 3 2 2 2)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F20",
   "case_id": "FG-F20",
   "kind": "task_score",
   "title": "Texas SMB: letter, then 'email it'",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.14,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F21",
   "case_id": "FG-F21",
   "kind": "session_score",
   "title": "Employee 44: sign tonight? (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.71,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 4 3 4 4 3 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F21",
   "case_id": "FG-F21",
   "kind": "task_score",
   "title": "Employee 44: sign tonight? (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F22",
   "case_id": "FG-F22",
   "kind": "session_score",
   "title": "Consultant: NDA on phone, Skip, Approve, follow-up",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 1.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 2 2 2 1 2 1)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F22",
   "case_id": "FG-F22",
   "kind": "task_score",
   "title": "Consultant: NDA on phone, Skip, Approve, follow-up",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F23",
   "case_id": "FG-F23",
   "kind": "session_score",
   "title": "Founder: follow-up dilution math (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 5 3 5 4 3 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F23",
   "case_id": "FG-F23",
   "kind": "task_score",
   "title": "Founder: follow-up dilution math (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F24",
   "case_id": "FG-F24",
   "kind": "session_score",
   "title": "England divorce: re-test of session 5",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.71,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (rubric scored from session notes: 4 3 4 4 3 4)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F24",
   "case_id": "FG-F24",
   "kind": "task_score",
   "title": "England divorce: re-test of session 5",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.10 plus queued D and code fixes 28 to 38, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S8-24-baseline:FG-F8-24 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F8-24 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S8-24-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.99,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C11-v2.10:FG-F8-24 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F8-24 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C11-v2.10",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.32,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "prompt items alone estimated lower; pipeline items 28 to 31 carry most of the gain",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:ALL: all tasks",
   "case_id": null,
   "kind": "run_summary",
   "title": "ALL: all tasks",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.784",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "2 of 42 measured tasks self-serve.; 40 of 51 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: general chat",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: general chat",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.5",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 3 measured tasks self-serve.; 2 of 4 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: document edit and redline",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: document edit and redline",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.8",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 4 measured tasks self-serve.; 4 of 5 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: send and sign",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: send and sign",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 2 measured tasks self-serve. LOW-N.; 4 of 4 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: Vault summary and obligations",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: Vault summary and obligations",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 0 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: obligations and payments tracking",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: obligations and payments tracking",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 0 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: contract and term simulator",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: contract and term simulator",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 0 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: counterparty markup responses",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: counterparty markup responses",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 0 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: editor chat edits and Responses",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: editor chat edits and Responses",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 0 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: onboarding",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: onboarding",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 1 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: employment",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: employment",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "1 of 4 measured tasks self-serve.; 4 of 4 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: disputes",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: disputes",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 4 measured tasks self-serve.; 4 of 4 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: family",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: family",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.8",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 4 measured tasks self-serve.; 4 of 5 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: fundraising",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: fundraising",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.833",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "1 of 5 measured tasks self-serve.; 5 of 6 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: commercial and DPA",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: commercial and DPA",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 6 measured tasks self-serve.; 6 of 6 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: customer app A (bank and regulated DPAs)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app A (bank and regulated DPAs)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.833",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 5 measured tasks self-serve.; 5 of 6 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: customer app B (sports sponsorship data room)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app B (sports sponsorship data room)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.667",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 2 measured tasks self-serve. LOW-N.; 2 of 3 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: customer app C (UK fundraising)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app C (UK fundraising)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 3 measured tasks self-serve.; 0 of 3 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:UC: customer app D (venture fund paperwork)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app D (venture fund paperwork)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 0 measured tasks self-serve. LOW-N.; 0 of 0 estimated tasks self-serve after queued fixes and v2.10, not re-run. LOW-N.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:SEG: individuals",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: individuals",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.889",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "1 of 8 measured tasks self-serve.; 8 of 9 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:SEG: founders",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: founders",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.824",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "1 of 12 measured tasks self-serve.; 14 of 17 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:SEG: SMBs",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: SMBs",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.889",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 8 measured tasks self-serve.; 8 of 9 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:SS-C12:SEG: legal teams",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: legal teams",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.625",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0 of 14 measured tasks self-serve.; 10 of 16 estimated tasks self-serve after queued fixes and v2.10, not re-run.",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F1",
   "case_id": "FG-F1",
   "kind": "session_score",
   "title": "Series A MFN waiver: edit my docx, redline, sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 4.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 3, investment 2. feedback/founders-smb.md WTP: weekly in a raise, daily only with signing and obligations; docs lost on Continue and follow-ups",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F10",
   "case_id": "FG-F10",
   "kind": "session_score",
   "title": "Texas SMB: bounced $3,200 check, write the letter",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 2, variable reward 3, investment 1. occasional use; questions before the draft; letter lost on follow-up",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F11",
   "case_id": "FG-F11",
   "kind": "session_score",
   "title": "England divorce: no-fault, timing, court (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 3, variable reward 3, investment 1. episodic; pays per solicitor-checked step",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F12",
   "case_id": "FG-F12",
   "kind": "session_score",
   "title": "California consultant: SOW red flags (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 3, investment 1. weekly; strong NDA verdicts; nothing stored for next time",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F13",
   "case_id": "FG-F13",
   "kind": "session_score",
   "title": "Series A founder: MFN waiver on phone, Continue path",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 4.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 3, investment 2. feedback/founders-smb.md WTP: weekly in a raise, daily only with signing and obligations; docs lost on Continue and follow-ups",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F14",
   "case_id": "FG-F14",
   "kind": "session_score",
   "title": "Texas SMB: 'write it now' with all facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 2, variable reward 3, investment 1. occasional use; questions before the draft; letter lost on follow-up",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F15",
   "case_id": "FG-F15",
   "kind": "session_score",
   "title": "Founder: 'just give me the number' (SAFE %)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 2, action 3, variable reward 3, investment 2. weekly during the raise; number given but behind a scope card; no running cap table",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F16",
   "case_id": "FG-F16",
   "kind": "session_score",
   "title": "Consultant: NDA with 'no questions' (desktop)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 3, investment 1. weekly; strong NDA verdicts; nothing stored for next time",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F17",
   "case_id": "FG-F17",
   "kind": "session_score",
   "title": "England: threat to move children to Glasgow (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 3, variable reward 3, investment 1. episodic; pays per solicitor-checked step",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F18",
   "case_id": "FG-F18",
   "kind": "session_score",
   "title": "San Jose employee: final pay on next payday",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 0, action 3, variable reward 3, investment 0. one-off crisis; $79 one-off",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F19",
   "case_id": "FG-F19",
   "kind": "session_score",
   "title": "Series A founder: clean doc plus investor emails",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 4.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 3, investment 2. feedback/founders-smb.md WTP: weekly in a raise, daily only with signing and obligations; docs lost on Continue and follow-ups",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F2",
   "case_id": "FG-F2",
   "kind": "session_score",
   "title": "First-time founder: SAFE dilution on a phone",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 2, action 3, variable reward 3, investment 2. weekly during the raise; number given but behind a scope card; no running cap table",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F20",
   "case_id": "FG-F20",
   "kind": "session_score",
   "title": "Texas SMB: letter, then 'email it'",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 2, variable reward 3, investment 1. occasional use; questions before the draft; letter lost on follow-up",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F21",
   "case_id": "FG-F21",
   "kind": "session_score",
   "title": "Employee 44: sign tonight? (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 0, action 3, variable reward 3, investment 0. one-off crisis; $79 one-off",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F22",
   "case_id": "FG-F22",
   "kind": "session_score",
   "title": "Consultant: NDA on phone, Skip, Approve, follow-up",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 3, investment 1. weekly; strong NDA verdicts; nothing stored for next time",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F23",
   "case_id": "FG-F23",
   "kind": "session_score",
   "title": "Founder: follow-up dilution math (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing maths",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 2, action 3, variable reward 3, investment 2. weekly during the raise; number given but behind a scope card; no running cap table",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F24",
   "case_id": "FG-F24",
   "kind": "session_score",
   "title": "England divorce: re-test of session 5",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 3, variable reward 3, investment 1. episodic; pays per solicitor-checked step",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F3",
   "case_id": "FG-F3",
   "kind": "session_score",
   "title": "Texas unpaid invoice: demand letter now",
   "source": "focus group, anonymised real inputs",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 2, variable reward 3, investment 1. occasional use; questions before the draft; letter lost on follow-up",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F4",
   "case_id": "FG-F4",
   "kind": "session_score",
   "title": "Solo consultant: NDA top risks, sign today",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 3, investment 1. weekly; strong NDA verdicts; nothing stored for next time",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F5",
   "case_id": "FG-F5",
   "kind": "session_score",
   "title": "Divorce in England, two children, joint home",
   "source": "focus group, anonymised real inputs",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 1, action 3, variable reward 3, investment 1. episodic; pays per solicitor-checked step",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F6",
   "case_id": "FG-F6",
   "kind": "session_score",
   "title": "Employee in California, HR meeting tomorrow",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 0, action 3, variable reward 3, investment 0. one-off crisis; $79 one-off",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F7",
   "case_id": "FG-F7",
   "kind": "session_score",
   "title": "Angel wants a board seat on a SAFE",
   "source": "focus group, anonymised real inputs",
   "use_case": "Financing negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 2, action 3, variable reward 3, investment 2. weekly during the raise; number given but behind a scope card; no running cap table",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F8",
   "case_id": "FG-F8",
   "kind": "session_score",
   "title": "Series A founder: one-shot MFN edit, clean, redline, e-sign (simulated)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 4.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 3, investment 2. feedback/founders-smb.md WTP: weekly in a raise, daily only with signing and obligations; docs lost on Continue and follow-ups",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-F9",
   "case_id": "FG-F9",
   "kind": "session_score",
   "title": "Employee 44: severance release, 7 days to sign (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "1.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 0, action 3, variable reward 3, investment 0. one-off crisis; $79 one-off",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L1",
   "case_id": "FG-L1",
   "kind": "session_score",
   "title": "Bank counterparty: explain the regulatory driver",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 4, investment 2. feedback/legal-teams.md FINAL persona scores: delight 7; 55 to 80 s first view",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L10",
   "case_id": "FG-L10",
   "kind": "session_score",
   "title": "EIS and SEIS connected persons",
   "source": "focus group, anonymised real inputs",
   "use_case": "Tax advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 4, investment 2. delight 7; false review label undercuts the reward",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L11",
   "case_id": "FG-L11",
   "kind": "session_score",
   "title": "One banking breach notice position across apps",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 4, action 3, variable reward 4, investment 3. delight 7; every addendum round; playbook is stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L12",
   "case_id": "FG-L12",
   "kind": "session_score",
   "title": "UK fundraising: financial promotion restriction and exemptions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 4, investment 2. delight 7; false review label undercuts the reward",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L13",
   "case_id": "FG-L13",
   "kind": "session_score",
   "title": "Controller or processor before recommending a DPA",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.25",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 4, investment 3. delight 8; data room and reports are stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L14",
   "case_id": "FG-L14",
   "kind": "session_score",
   "title": "Penalty check on repayment clauses by governing law",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "CH;SG;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.25",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 4, investment 3. delight 8; data room and reports are stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L15",
   "case_id": "FG-L15",
   "kind": "session_score",
   "title": "California work made for hire cite",
   "source": "focus group, anonymised real inputs",
   "use_case": "Drafting",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 4, investment 2. feedback/legal-teams.md FINAL persona scores: delight 7; 55 to 80 s first view",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L16",
   "case_id": "FG-L16",
   "kind": "session_score",
   "title": "UK NSI thresholds with cites",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 4, investment 2. feedback/legal-teams.md FINAL persona scores: delight 7; 55 to 80 s first view",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L2",
   "case_id": "FG-L2",
   "kind": "session_score",
   "title": "Answer first, then 1 or 2 facts",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 4, investment 2. feedback/legal-teams.md FINAL persona scores: delight 7; 55 to 80 s first view",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L3",
   "case_id": "FG-L3",
   "kind": "session_score",
   "title": "Lawyer quote needs user consent",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 4, investment 2. feedback/legal-teams.md FINAL persona scores: delight 7; 55 to 80 s first view",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L4",
   "case_id": "FG-L4",
   "kind": "session_score",
   "title": "First view under 10 seconds",
   "source": "focus group, anonymised real inputs",
   "use_case": "Latency",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "2.75",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 2, variable reward 4, investment 2. feedback/legal-teams.md FINAL persona scores: delight 7; 55 to 80 s first view",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L5",
   "case_id": "FG-L5",
   "kind": "session_score",
   "title": "Governing law in a data transfer context",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract redline",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 4, action 3, variable reward 4, investment 3. delight 7; every addendum round; playbook is stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L6",
   "case_id": "FG-L6",
   "kind": "session_score",
   "title": "Processor breach notice for a bank: 24 hours, not 72",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract redline",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 4, action 3, variable reward 4, investment 3. delight 7; every addendum round; playbook is stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L7",
   "case_id": "FG-L7",
   "kind": "session_score",
   "title": "Critical or important function: DORA Art. 30 mapping",
   "source": "focus group, anonymised real inputs",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 4, action 3, variable reward 4, investment 3. delight 7; every addendum round; playbook is stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L8",
   "case_id": "FG-L8",
   "kind": "session_score",
   "title": "After flagging Privacy Shield, name the current route",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "EU;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.5",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 4, action 3, variable reward 4, investment 3. delight 7; every addendum round; playbook is stored investment",
   "segment": null
  },
  {
   "row_id": "scores:HOOK-C12:FG-L9",
   "case_id": "FG-L9",
   "kind": "session_score",
   "title": "Honest review label until a lawyer signs off",
   "source": "focus group, anonymised real inputs",
   "use_case": "Behavior",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HOOK-C12",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": "hooked",
   "metric_value": "3.0",
   "basis": "measured",
   "label": "MEASURED (focus group persona notes, scored against the Hooked sub-score)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "trigger 3, action 3, variable reward 4, investment 2. delight 7; false review label undercuts the reward",
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F25",
   "case_id": "FG-F25",
   "kind": "session_score",
   "title": "Series A founder: .docx redline plus cover email, no questions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 3.0,
   "e2e_wow": 1.0,
   "task_score": 1.57,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Upload not attached (.docx), asked for the waiver it was given, 9 then 10 facts missing for a cover email.",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F25",
   "case_id": "FG-F25",
   "kind": "task_score",
   "title": "Series A founder: .docx redline plus cover email, no questions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F26",
   "case_id": "FG-F26",
   "kind": "session_score",
   "title": "CA employee, 52, group layoff: release in 7 days (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 3.0,
   "polish": 3.0,
   "e2e_wow": 2.0,
   "task_score": 3.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Legal answer correct; Emails sheet took over the screen and blocked the composer for 30 s.",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F26",
   "case_id": "FG-F26",
   "kind": "task_score",
   "title": "CA employee, 52, group layoff: release in 7 days (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.29,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F27",
   "case_id": "FG-F27",
   "kind": "session_score",
   "title": "Series A founder: .docx with wait, signing email",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 1.0,
   "polish": 1.0,
   "e2e_wow": 1.0,
   "task_score": 1.43,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Upload not attached; Thai spam tokens and \"ETwitter\" in thinking; misread individual signatures; stalled with Documents empty.",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F27",
   "case_id": "FG-F27",
   "kind": "task_score",
   "title": "Series A founder: .docx with wait, signing email",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F28",
   "case_id": "FG-F28",
   "kind": "session_score",
   "title": "Consultant: .txt NDA, top 3 risks, no questions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 4.0,
   "grounding": 3.0,
   "specificity": 4.0,
   "actionability": 4.0,
   "polish": 3.0,
   "e2e_wow": 4.0,
   "task_score": 3.71,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Strong verdict; lawyer review default with US$1,650 cap and 0.4166666666666667h; first text 35 s.",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F28",
   "case_id": "FG-F28",
   "kind": "task_score",
   "title": "Consultant: .txt NDA, top 3 risks, no questions",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.43,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F29",
   "case_id": "FG-F29",
   "kind": "session_score",
   "title": "Series A founder: .docx upload as anonymous user hits a Vault sign-in wall",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 1.0,
   "grounding": 1.0,
   "specificity": 1.0,
   "actionability": 1.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 1.14,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Full-page Vault sign-in gate after upload; composer and typed request lost; no answer in 150 s.",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F29",
   "case_id": "FG-F29",
   "kind": "task_score",
   "title": "Series A founder: .docx upload as anonymous user hits a Vault sign-in wall",
   "source": "focus group, anonymised real inputs",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 3.71,
   "weight": 4.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F30",
   "case_id": "FG-F30",
   "kind": "session_score",
   "title": "CA employee, 44: Skip, Approve, follow-up (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 5.0,
   "grounding": 4.0,
   "specificity": 5.0,
   "actionability": 5.0,
   "polish": 4.0,
   "e2e_wow": 4.0,
   "task_score": 4.43,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Complete answer with 626(f) and 12964.5 cites, follow-up kept files; status stuck after Approve; first text 53 s.",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F30",
   "case_id": "FG-F30",
   "kind": "task_score",
   "title": "CA employee, 44: Skip, Approve, follow-up (mobile)",
   "source": "focus group, anonymised real inputs",
   "use_case": "Employment advice",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.57,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F31",
   "case_id": "FG-F31",
   "kind": "session_score",
   "title": "Consultant: .txt NDA on phone, Skip and follow-up",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": 2.0,
   "grounding": 2.0,
   "specificity": 2.0,
   "actionability": 1.0,
   "polish": 2.0,
   "e2e_wow": 1.0,
   "task_score": 1.57,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (focus group session notes, scored against rubric)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Final said it could not see the terms; follow-up showed 2 files attached but asked for the text; past target date; \"founder intake\".",
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F31",
   "case_id": "FG-F31",
   "kind": "task_score",
   "title": "Consultant: .txt NDA on phone, Skip and follow-up",
   "source": "focus group, anonymised real inputs",
   "use_case": "Contract review",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.0,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE (v2.11 plus C13 code items 39 to 43, not built, not re-run)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:FG-F-S25-31-baseline:FG-F25-31 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F25-31 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "FG-F-S25-31-baseline",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 2.41,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:GC-C13-v2.11:FG-F25-31 avg",
   "case_id": null,
   "kind": "run_summary",
   "title": "FG-F25-31 avg",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "GC-C13-v2.11",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": 4.14,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "estimated",
   "label": "ESTIMATE",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: general chat",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: general chat",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.5",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: answers reached (FG-F23, F18), scope cards and stalls elsewhere; outcome evidence: answers reached (FG-F23, F18), scope cards and stalls elsewhere",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: document edit and redline",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: document edit and redline",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.833",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: redline produced (QA pass 4, S8), .docx not attached (S25, S27), no clean copy; outcome evidence: redline produced (QA pass 4, S8), .docx not attached (S25, S27), no clean copy",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: send and sign",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: send and sign",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.833",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: no send or sign control (QA pass 3 and 4; S29 blocked); outcome evidence: no send or sign control (QA pass 3 and 4; S29 blocked)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: Vault summary and obligations",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: Vault summary and obligations",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: /vault 404 on Pascal-1 (QA pass 3); customer app D extraction works; outcome evidence: /vault 404 on Pascal-1 (QA pass 3); customer app D extraction works",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: obligations and payments tracking",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: obligations and payments tracking",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: Pascal-1 live run CU R4 (10:35 AM PT) gave a generic checklist, no CTA to the Obligations Monitor; customer app D tracker partial; outcome evidence: Pascal-1 live run CU R4 (10:35 AM PT) gave a generic checklist, no CTA to the Obligations Monitor; customer app D tracker partial",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: contract and term simulator",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: contract and term simulator",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: not yet tested; outcome evidence: not yet tested",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: counterparty markup responses",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: counterparty markup responses",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: customer-derived counters produced, not bank-safe (Legal S2); outcome evidence: customer-derived counters produced, not bank-safe (Legal S2)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: editor chat edits and Responses",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: editor chat edits and Responses",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: not yet tested; outcome evidence: not yet tested",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: onboarding",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: onboarding",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: profile modal fixed; anonymous upload wall (S29); outcome evidence: profile modal fixed; anonymous upload wall (S29)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: employment",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: employment",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: complete answer and scripts (S30); nothing sent; outcome evidence: complete answer and scripts (S30); nothing sent",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: disputes",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: disputes",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: letters drafted; \"email it\" lost the letter (S20); outcome evidence: letters drafted; \"email it\" lost the letter (S20)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: family",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: family",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.8",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: advice and checklist; no filing step; outcome evidence: advice and checklist; no filing step",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: fundraising",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: fundraising",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.833",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: numbers and drafts; send term sheet not reached (real-vs-canned.md); outcome evidence: numbers and drafts; send term sheet not reached (real-vs-canned.md)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: commercial and DPA",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: commercial and DPA",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "1.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: NDA verdicts (S28); file not read on Skip (S31); outcome evidence: NDA verdicts (S28); file not read on Skip (S31)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: customer app A (bank and regulated DPAs)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app A (bank and regulated DPAs)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.833",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: review, package and DOCX exports; counters not send-ready (Legal S2); outcome evidence: review, package and DOCX exports; counters not send-ready (Legal S2)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: customer app B (sports sponsorship data room)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app B (sports sponsorship data room)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.667",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: report and instruction email; stale report, no export (Legal S2, S4); outcome evidence: report and instruction email; stale report, no export (Legal S2, S4)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: customer app C (UK fundraising)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app C (UK fundraising)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: pipeline real; false review label (Legal S3); outcome evidence: pipeline real; false review label (Legal S3)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:UC: customer app D (venture fund paperwork)",
   "case_id": null,
   "kind": "run_summary",
   "title": "UC: customer app D (venture fund paperwork)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: draft notice to counterparty; Mark as sent disabled (customer-derived screenshots); outcome evidence: draft notice to counterparty; Mark as sent disabled (customer-derived screenshots)",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:SEG: individuals",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: individuals",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.909",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: see outcome-breaks.md; outcome evidence: see outcome-breaks.md",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:SEG: founders",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: founders",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.8",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: see outcome-breaks.md; outcome evidence: see outcome-breaks.md",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:SEG: SMBs",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: SMBs",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.909",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: see outcome-breaks.md; outcome evidence: see outcome-breaks.md",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:SEG: legal teams",
   "case_id": null,
   "kind": "run_summary",
   "title": "SEG: legal teams",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.625",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: see outcome-breaks.md; outcome evidence: see outcome-breaks.md",
   "segment": null
  },
  {
   "row_id": "scores:SS-C14:ALL: all tasks",
   "case_id": null,
   "kind": "run_summary",
   "title": "ALL: all tasks",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "SS-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_serve",
   "metric_value": "0.793",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "outcome evidence: see outcome-breaks.md; outcome evidence: see outcome-breaks.md",
   "segment": null
  },
  {
   "row_id": "scores:CU-C14:CU: NDA to customer-derived",
   "case_id": null,
   "kind": "run_summary",
   "title": "CU: NDA to [name]",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "CU-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "outcome",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (computer-use run 10:35 AM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "After Send redirected to /pascal?q= and showed 'Sign in to open your company's private Vault' on the pascal-1 preview host and staging; request lost, nothing drafted, no text. Screenshots 029bbb9afdcb36482481bd9801ba408cc3513486c0170376a33eddaa8e1cba2a.png, 18005f8c143580c0c890e32edca42068425ef9ba8e800a117fdd67e3aa968e70.png",
   "segment": null
  },
  {
   "row_id": "scores:CU-C14:CU: seed term sheet",
   "case_id": null,
   "kind": "run_summary",
   "title": "CU: seed term sheet",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "CU-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "outcome",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (computer-use run 10:35 AM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Same Vault sign-in wall after Send; button spun about 6 s with no message; nothing drafted. Screenshots 029bbb9afdcb36482481bd9801ba408cc3513486c0170376a33eddaa8e1cba2a.png, 18005f8c143580c0c890e32edca42068425ef9ba8e800a117fdd67e3aa968e70.png",
   "segment": null
  },
  {
   "row_id": "scores:CU-C14:CU: UAE contractor with no trade licence",
   "case_id": null,
   "kind": "run_summary",
   "title": "CU: UAE contractor with no trade licence",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "CU-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "outcome",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (computer-use run 10:35 AM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Routed to Contract Redliner instead of drafting; recommended a lawyer; said it can't send documents or obtain the signature; no draft, Documents empty; 3 key questions collapsed under Optional refinements; task later showed another task's payments answer (crossed streams, P0). Screenshots d1f04855dad71e9b5b286f10b3057eab591ccc78f693496c5308b07dd5c1aaa5.png, a19ec1d8f4611aba9cd2f508c3f48f6bb852bf06",
   "segment": null
  },
  {
   "row_id": "scores:CU-C14:CU: when do I owe payments",
   "case_id": null,
   "kind": "run_summary",
   "title": "CU: when do I owe payments",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "CU-C14",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "outcome",
   "metric_value": null,
   "basis": "measured",
   "label": "MEASURED (computer-use run 10:35 AM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Generic checklist; no Upload, Use example data or Obligations Monitor link; cited a $4,000 contractor fee from a different chat (cross-chat leak, P0); Overall confidence: High with no contracts; raw ** markdown. Screenshots 91a07a6e0f4d4a2ef6787786167d01679cd0390fb436fe880caba2bcda408826.png, 14478feb992de998a63ea828fd6f330a6819cea6f073d670eb768ecc784f72bc.png",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-C1:Routing eval: drafting, main as-is (pascal-1)",
   "case_id": null,
   "kind": "run_summary",
   "title": "Routing eval: drafting, main as-is (pascal-1)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-C1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "drafting_pass",
   "metric_value": "0/27",
   "basis": "measured",
   "label": "MEASURED (harness 11:20 to 11:45 AM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "0% overall; mfn_waiver 0/8, side_letter 0/2, nda 0/4, advisor 0/2, safe 0/2, uae_nda 0/4, uae_services 0/5. MFN drafts a SAFE (parse.ts:42), 2 year term becomes $2M (parse.ts:53-54), request text never reaches the model (signing.functions.ts:36-48), other becomes seed term sheet (signing.functions.ts:46), My company party (signing.functions.ts:25)",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-C1:Routing eval: routing, main as-is",
   "case_id": null,
   "kind": "run_summary",
   "title": "Routing eval: routing, main as-is",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-C1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "routing_pass",
   "metric_value": "7/27",
   "basis": "measured",
   "label": "MEASURED (route.mts on main 9c067a2, no model)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "26% on hill-climber and qa-pod cases; 11/53 (21%) including 26 main-worker cases",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-C1:Routing eval: drafting, main as-is (gpt-6-astra, production ",
   "case_id": null,
   "kind": "run_summary",
   "title": "Routing eval: drafting, main as-is (gpt-6-astra, production model)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-C1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "drafting_pass",
   "metric_value": "0/4",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "RC-001, 002, 017, 019; same failures as pascal-1, so model choice is not the cause",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-C1:Routing eval: drafting, owner prompt shape only (pascal-1)",
   "case_id": null,
   "kind": "run_summary",
   "title": "Routing eval: drafting, owner prompt shape only (pascal-1)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-C1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "drafting_pass",
   "metric_value": "7/10",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "verbatim request, doc type label, no My company; repo DRAFT_INSTRUCTIONS; fails ADGM law basis, dropped term, invented total",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-C1:Routing eval: drafting, proposed fix v1 (pascal-1)",
   "case_id": null,
   "kind": "run_summary",
   "title": "Routing eval: drafting, proposed fix v1 (pascal-1)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-C1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "drafting_pass",
   "metric_value": "26/27",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "96%; only RC-013 vesting cliff dropped; v2 addendum rerun 4/4 (RC-001, 013, 019, 027). Patch: evals/routing/proposed/0001-draft-instructions-doc-type-and-facts.patch, not applied",
   "segment": null
  },
  {
   "row_id": "scores:HARDEN-C1:Prompt hardening pass: 55 prompts on main@855a5ea7, patches ",
   "case_id": null,
   "kind": "run_summary",
   "title": "Prompt hardening pass: 55 prompts on main@855a5ea7, patches 01 to 10 (proposal, not applied)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "HARDEN-C1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "prompt_lint_failures",
   "metric_value": "71 -> 0",
   "basis": "measured",
   "label": "MEASURED",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "Static lint, 55 prompts: clean 10 -> 55. Deterministic research rerun unchanged: research-citations 15/15, citation-hallucination 20/20, 0 fabricated. Intent 97-case live run pending (no session: PC unset, eval account unknown); upl-safety and model research pending (staging). No model score claimed for patched prompts.",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-RL1:Routing eval: realistic RL set, main regex router",
   "case_id": null,
   "kind": "run_summary",
   "title": "Routing eval: realistic RL set, main regex router",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-RL1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "amounts_ok",
   "metric_value": "41/102",
   "basis": "measured",
   "label": "MEASURED (route.mts on main 9c067a2, no model, 31 Oct 2026 2:20 PM PT)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "17%; by segment yc_seed 5/30, vc 0/6, sv_inhouse 2/21, sports_events 4/13, smb_family 2/11, signing_recipient 0/3, cross_border 1/7, adversarial 3/11; multi-turn 3/20; easy 1/8, medium 8/47, hard 8/47; train 13/73, held-out 4/29; top misroutes: deal_draft to vault_answer 27, vault_answer to doc_revise 14; mfn_waiver 0/5 (routed as a new SAFE); durations and dates parsed as money in employment, Del",
   "segment": null
  },
  {
   "row_id": "scores:ROUTE-RL1:Intent eval: oracle self-check, all 199 cases (consistency, ",
   "case_id": null,
   "kind": "run_summary",
   "title": "Intent eval: oracle self-check, all 199 cases (consistency, not a model score)",
   "source": "run summary",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": "ROUTE-RL1",
   "split": null,
   "sample": null,
   "scale": "1-5",
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "self_check_oracle_pass",
   "metric_value": "199/199",
   "basis": "measured",
   "label": "MEASURED (offline self-check, no network)",
   "judge": null,
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/scores.csv",
   "notes": "lint errors only RC-005 and RC-020 (earlier hill-climber lines, untouched); summary evals/intent/runs/2026-10-31-self-check-final.summary.md",
   "segment": null
  },
  {
   "row_id": "route-only:RC-001",
   "case_id": "RC-001",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-001",
   "case_id": "RC-001",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-002",
   "case_id": "RC-002",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-002",
   "case_id": "RC-002",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-003",
   "case_id": "RC-003",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-003",
   "case_id": "RC-003",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-004",
   "case_id": "RC-004",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-004",
   "case_id": "RC-004",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-005",
   "case_id": "RC-005",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument safe; agent_ok False; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-005",
   "case_id": "RC-005",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-006",
   "case_id": "RC-006",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-006",
   "case_id": "RC-006",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-007",
   "case_id": "RC-007",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-007",
   "case_id": "RC-007",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-008",
   "case_id": "RC-008",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-008",
   "case_id": "RC-008",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-009",
   "case_id": "RC-009",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-009",
   "case_id": "RC-009",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-010",
   "case_id": "RC-010",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-010",
   "case_id": "RC-010",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-011",
   "case_id": "RC-011",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument nda; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-011",
   "case_id": "RC-011",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-012",
   "case_id": "RC-012",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-012",
   "case_id": "RC-012",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-013",
   "case_id": "RC-013",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "advisor",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-013",
   "case_id": "RC-013",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "advisor",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction False; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-014",
   "case_id": "RC-014",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-014",
   "case_id": "RC-014",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms False; number missing: 500,000; number missing: 8,000,000",
   "segment": null
  },
  {
   "row_id": "route-only:RC-015",
   "case_id": "RC-015",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-015",
   "case_id": "RC-015",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction False; parties True; terms False; number missing: 250,000; number missing: 10,000,000",
   "segment": null
  },
  {
   "row_id": "route-only:RC-016",
   "case_id": "RC-016",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-016",
   "case_id": "RC-016",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-017",
   "case_id": "RC-017",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument nda; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-017",
   "case_id": "RC-017",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-018",
   "case_id": "RC-018",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-018",
   "case_id": "RC-018",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-019",
   "case_id": "RC-019",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-019",
   "case_id": "RC-019",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms False; number missing: 15,000",
   "segment": null
  },
  {
   "row_id": "route-only:RC-020",
   "case_id": "RC-020",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-020",
   "case_id": "RC-020",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms False; number missing: 120,000; number missing: 60,000",
   "segment": null
  },
  {
   "row_id": "route-only:RC-021",
   "case_id": "RC-021",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-021",
   "case_id": "RC-021",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms False; number missing: 40,000",
   "segment": null
  },
  {
   "row_id": "route-only:RC-022",
   "case_id": "RC-022",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-022",
   "case_id": "RC-022",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms False; number missing: 900",
   "segment": null
  },
  {
   "row_id": "route-only:RC-023",
   "case_id": "RC-023",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-023",
   "case_id": "RC-023",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task False; doc_type True; jurisdiction False; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-024",
   "case_id": "RC-024",
   "kind": "routing_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:RC-024",
   "case_id": "RC-024",
   "kind": "intent_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties True; terms False; number missing: 22,000",
   "segment": null
  },
  {
   "row_id": "route-only:RC-025",
   "case_id": "RC-025",
   "kind": "routing_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-025",
   "case_id": "RC-025",
   "kind": "intent_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-026",
   "case_id": "RC-026",
   "kind": "routing_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-026",
   "case_id": "RC-026",
   "kind": "intent_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:RC-027",
   "case_id": "RC-027",
   "kind": "routing_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:RC-027",
   "case_id": "RC-027",
   "kind": "intent_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-001",
   "case_id": "MW-001",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-001",
   "case_id": "MW-001",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-002",
   "case_id": "MW-002",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-002",
   "case_id": "MW-002",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-003",
   "case_id": "MW-003",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-003",
   "case_id": "MW-003",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-004",
   "case_id": "MW-004",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-004",
   "case_id": "MW-004",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-005",
   "case_id": "MW-005",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-005",
   "case_id": "MW-005",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-006",
   "case_id": "MW-006",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "nda_mutual",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-006",
   "case_id": "MW-006",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "nda_mutual",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type False; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-007",
   "case_id": "MW-007",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "nda_one_way",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-007",
   "case_id": "MW-007",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "nda_one_way",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-008",
   "case_id": "MW-008",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "board_resolution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "board_resolution",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-008",
   "case_id": "MW-008",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "board_resolution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "board_resolution",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-009",
   "case_id": "MW-009",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "shareholder_resolution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "shareholder_resolution",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-009",
   "case_id": "MW-009",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "shareholder_resolution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "shareholder_resolution",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-010",
   "case_id": "MW-010",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "consent_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "consent_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-010",
   "case_id": "MW-010",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "consent_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "consent_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-011",
   "case_id": "MW-011",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "consent_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "consent_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-011",
   "case_id": "MW-011",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "consent_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "consent_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-012",
   "case_id": "MW-012",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "amendment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "amendment",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-012",
   "case_id": "MW-012",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "amendment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "amendment",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-013",
   "case_id": "MW-013",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "termination_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "termination_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-013",
   "case_id": "MW-013",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "termination_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "termination_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-014",
   "case_id": "MW-014",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "ip_assignment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "ip_assignment",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-014",
   "case_id": "MW-014",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "ip_assignment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "ip_assignment",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-015",
   "case_id": "MW-015",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "offer_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "offer_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-015",
   "case_id": "MW-015",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "offer_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "offer_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-016",
   "case_id": "MW-016",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "employment_agreement",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "employment_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-016",
   "case_id": "MW-016",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "employment_agreement",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "employment_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-017",
   "case_id": "MW-017",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "dpa",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "dpa",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-017",
   "case_id": "MW-017",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "dpa",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "dpa",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-018",
   "case_id": "MW-018",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "sow",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "sow",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-018",
   "case_id": "MW-018",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "sow",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "sow",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-019",
   "case_id": "MW-019",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "convertible_note",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "convertible_note",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:MW-019",
   "case_id": "MW-019",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "convertible_note",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "convertible_note",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-020",
   "case_id": "MW-020",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "term_sheet",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument term_sheet; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-020",
   "case_id": "MW-020",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "term_sheet",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-021",
   "case_id": "MW-021",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "term_sheet",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument term_sheet; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-021",
   "case_id": "MW-021",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "term_sheet",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-022",
   "case_id": "MW-022",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "safe_post_money_cap",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe_post_money_cap",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-022",
   "case_id": "MW-022",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "safe_post_money_cap",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe_post_money_cap",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-023",
   "case_id": "MW-023",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "unknown",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "unknown",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-023",
   "case_id": "MW-023",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "unknown",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "unknown",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task False; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-024",
   "case_id": "MW-024",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "unknown",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "unknown",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-024",
   "case_id": "MW-024",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "unknown",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "unknown",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-025",
   "case_id": "MW-025",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "safe_post_money_cap",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe_post_money_cap",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-025",
   "case_id": "MW-025",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "safe_post_money_cap",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe_post_money_cap",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type False; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:MW-026",
   "case_id": "MW-026",
   "kind": "routing_case",
   "title": null,
   "source": "main-worker",
   "use_case": "nda_mutual",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:MW-026",
   "case_id": "MW-026",
   "kind": "intent_case",
   "title": null,
   "source": "main-worker",
   "use_case": "nda_mutual",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type False; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:IR-001",
   "case_id": "IR-001",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-001",
   "case_id": "IR-001",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; deliverables missing: ['envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-002",
   "case_id": "IR-002",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-002",
   "case_id": "IR-002",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; deliverables missing: ['envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-003",
   "case_id": "IR-003",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-003",
   "case_id": "IR-003",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; number missing: 250,000; number missing: 8,000,000; file safe-harborlane-northwind-with-mfn.md as source_instrument not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-004",
   "case_id": "IR-004",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-004",
   "case_id": "IR-004",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-005",
   "case_id": "IR-005",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-005",
   "case_id": "IR-005",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-006",
   "case_id": "IR-006",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "question",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument safe; agent_ok False; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-006",
   "case_id": "IR-006",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "question",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; number missing: 300,000",
   "segment": null
  },
  {
   "row_id": "route-only:IR-007",
   "case_id": "IR-007",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument safe; agent_ok True; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-007",
   "case_id": "IR-007",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; file nda-acme-consulting.txt as subject not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-008",
   "case_id": "IR-008",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-008",
   "case_id": "IR-008",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; deliverables missing: ['envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-009",
   "case_id": "IR-009",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-009",
   "case_id": "IR-009",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; number missing: 1,500,000",
   "segment": null
  },
  {
   "row_id": "route-only:IR-010",
   "case_id": "IR-010",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "demand_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path simulate; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-010",
   "case_id": "IR-010",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "demand_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; number missing: 8,400",
   "segment": null
  },
  {
   "row_id": "route-only:IR-011",
   "case_id": "IR-011",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument nda; agent_ok False; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-011",
   "case_id": "IR-011",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; number missing: 6,000,000; number missing: 100,000",
   "segment": null
  },
  {
   "row_id": "route-only:IR-012",
   "case_id": "IR-012",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-012",
   "case_id": "IR-012",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties True; terms False; number missing: 6,000,000; number missing: 100,000",
   "segment": null
  },
  {
   "row_id": "route-only:IR-013",
   "case_id": "IR-013",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "term_sheet",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument term_sheet; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-013",
   "case_id": "IR-013",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "term_sheet",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task False; doc_type True; jurisdiction None; parties None; terms False; number missing: 1,500,000; number missing: 12,000,000; file brightloop-standard-term-sheet.docx as template not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-014",
   "case_id": "IR-014",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument safe; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-014",
   "case_id": "IR-014",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task False; doc_type True; jurisdiction None; parties None; terms False; number missing: 1,500,000; number missing: 12,000,000; file safe-template-post-money-customer-derived.md as template not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-015",
   "case_id": "IR-015",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "dilution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument safe; agent_ok True; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-015",
   "case_id": "IR-015",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "dilution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['answer']; constraint no_questions != True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-016",
   "case_id": "IR-016",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "dilution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-016",
   "case_id": "IR-016",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "dilution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; constraint no_questions != True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-017",
   "case_id": "IR-017",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "dilution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok True; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-017",
   "case_id": "IR-017",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "dilution",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:IR-018",
   "case_id": "IR-018",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument nda; agent_ok False; instrument_ok False; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-018",
   "case_id": "IR-018",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; file nda-acme-consulting.txt as subject not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-019",
   "case_id": "IR-019",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-019",
   "case_id": "IR-019",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; file nda-acme-consulting.txt as carried_from_thread not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-020",
   "case_id": "IR-020",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument nda; agent_ok False; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-020",
   "case_id": "IR-020",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda_review",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task False; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['email_draft', 'answer']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-021",
   "case_id": "IR-021",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-021",
   "case_id": "IR-021",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['redline', 'clean']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-022",
   "case_id": "IR-022",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path draft_from_revision; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-022",
   "case_id": "IR-022",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['redline', 'clean', 'envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-023",
   "case_id": "IR-023",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-023",
   "case_id": "IR-023",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; date missing: 2026-12-31; deliverables missing: ['clean', 'envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-024",
   "case_id": "IR-024",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "sow",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "sow",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path draft_from_revision; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-024",
   "case_id": "IR-024",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "sow",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "sow",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['redline', 'clean', 'envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-025",
   "case_id": "IR-025",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "markup_response",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument nda; agent_ok False; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-025",
   "case_id": "IR-025",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "markup_response",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; file pasted_text as subject not used",
   "segment": null
  },
  {
   "row_id": "route-only:IR-026",
   "case_id": "IR-026",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-026",
   "case_id": "IR-026",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; artifact_ref Demand letter != L1; deliverables missing: ['email_compose']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-027",
   "case_id": "IR-027",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-027",
   "case_id": "IR-027",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; artifact_ref Demand letter != L1",
   "segment": null
  },
  {
   "row_id": "route-only:IR-028",
   "case_id": "IR-028",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-028",
   "case_id": "IR-028",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; artifact_ref mfn-waiver-brightloop.docx != P1; deliverables missing: ['clean', 'envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-029",
   "case_id": "IR-029",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-029",
   "case_id": "IR-029",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; artifact_ref MFN waiver, clean != R1",
   "segment": null
  },
  {
   "row_id": "route-only:IR-030",
   "case_id": "IR-030",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": "demand_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-030",
   "case_id": "IR-030",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": "demand_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction False; parties None; terms False; number missing: 3,200; constraint no_questions != True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-031",
   "case_id": "IR-031",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "divorce",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-031",
   "case_id": "IR-031",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "divorce",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:IR-032",
   "case_id": "IR-032",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "employment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-032",
   "case_id": "IR-032",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "employment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['email_compose']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-033",
   "case_id": "IR-033",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "consulting",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-033",
   "case_id": "IR-033",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "consulting",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; number missing: 8,000; date missing: 2026-11-10",
   "segment": null
  },
  {
   "row_id": "route-only:IR-034",
   "case_id": "IR-034",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": "demand_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-034",
   "case_id": "IR-034",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "demand_letter",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": "demand_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction True; parties None; terms False; date missing: 2026-08-15; date missing: 2026-11-10",
   "segment": null
  },
  {
   "row_id": "route-only:IR-035",
   "case_id": "IR-035",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-035",
   "case_id": "IR-035",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; date missing: 2026-11-06",
   "segment": null
  },
  {
   "row_id": "route-only:IR-036",
   "case_id": "IR-036",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-036",
   "case_id": "IR-036",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task False; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['clean', 'email_compose']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-037",
   "case_id": "IR-037",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "contractor_template",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path doc_revise; instrument other; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-037",
   "case_id": "IR-037",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "contractor_template",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type False; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:IR-038",
   "case_id": "IR-038",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "board_seat",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument safe; agent_ok True; instrument_ok False; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-038",
   "case_id": "IR-038",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "board_seat",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-039",
   "case_id": "IR-039",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "employment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok True; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-039",
   "case_id": "IR-039",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "employment",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-040",
   "case_id": "IR-040",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "employment_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "demand_letter",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok False",
   "segment": null
  },
  {
   "row_id": "intent:IR-040",
   "case_id": "IR-040",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "employment_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "demand_letter",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms True",
   "segment": null
  },
  {
   "row_id": "route-only:IR-041",
   "case_id": "IR-041",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "answer_then_draft",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path vault_answer; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-041",
   "case_id": "IR-041",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "answer_then_draft",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; deliverables missing: ['email_draft']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-042",
   "case_id": "IR-042",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "upload_gate",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path draft_from_revision; instrument other; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-042",
   "case_id": "IR-042",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "upload_gate",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type False; jurisdiction None; parties None; terms False; deliverables missing: ['redline', 'clean', 'envelope']",
   "segment": null
  },
  {
   "row_id": "route-only:IR-043",
   "case_id": "IR-043",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "ambiguous",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path deal_draft; instrument nda; agent_ok True; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-043",
   "case_id": "IR-043",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "ambiguous",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms None",
   "segment": null
  },
  {
   "row_id": "route-only:IR-044",
   "case_id": "IR-044",
   "kind": "routing_case",
   "title": null,
   "source": "focus-group",
   "use_case": "multi_intent",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "regex router (route.mts, no model)",
   "prompt_version": null,
   "run": "route-only-v2-97",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "route_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic",
   "run_date": "2026-10-31T12:16:00-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/runs/route-only-v2-97.json",
   "notes": "path draft_from_revision; instrument safe; agent_ok False; instrument_ok True; amounts_ok True",
   "segment": null
  },
  {
   "row_id": "intent:IR-044",
   "case_id": "IR-044",
   "kind": "intent_case",
   "title": null,
   "source": "focus-group",
   "use_case": "multi_intent",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": "LLM intent step (staging /api/intent/eval)",
   "prompt_version": null,
   "run": "harden-c1-before",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": false,
   "metric": "intent_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic (run_intent.py)",
   "run_date": "2026-10-31T14:11-07:00",
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/intent/runs/harden-c1-before.scores.jsonl",
   "notes": "task True; doc_type True; jurisdiction None; parties None; terms False; number missing: 8,000,000; deliverables missing: ['redline', 'clean', 'envelope']",
   "segment": null
  },
  {
   "row_id": "case:RL-001",
   "case_id": "RL-001",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-002",
   "case_id": "RL-002",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-003",
   "case_id": "RL-003",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-004",
   "case_id": "RL-004",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "mfn_waiver",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-005",
   "case_id": "RL-005",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-006",
   "case_id": "RL-006",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-007",
   "case_id": "RL-007",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-008",
   "case_id": "RL-008",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-009",
   "case_id": "RL-009",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "board_consent",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "board_consent",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-010",
   "case_id": "RL-010",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "board_consent",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "board_consent",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-011",
   "case_id": "RL-011",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "equity_83b",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-012",
   "case_id": "RL-012",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "equity_83b",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "election_83b",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-013",
   "case_id": "RL-013",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-014",
   "case_id": "RL-014",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-015",
   "case_id": "RL-015",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "contractor_ip",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-016",
   "case_id": "RL-016",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "contractor_ip",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": "ip_assignment",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-017",
   "case_id": "RL-017",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "delaware_ops",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-018",
   "case_id": "RL-018",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "delaware_ops",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-019",
   "case_id": "RL-019",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "pro_rata",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-020",
   "case_id": "RL-020",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "safe",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-021",
   "case_id": "RL-021",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "vc_term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-022",
   "case_id": "RL-022",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "vc_term_sheet",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "term_sheet",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-023",
   "case_id": "RL-023",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "vc_dd",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-024",
   "case_id": "RL-024",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "vc_dd",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-025",
   "case_id": "RL-025",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-026",
   "case_id": "RL-026",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "msa_dpa",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-027",
   "case_id": "RL-027",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "msa_dpa",
   "jurisdiction": "US-NY",
   "customer": null,
   "doc_type": "msa",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-028",
   "case_id": "RL-028",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "msa_dpa",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-029",
   "case_id": "RL-029",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "privacy",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-030",
   "case_id": "RL-030",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "privacy",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-031",
   "case_id": "RL-031",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_us",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-032",
   "case_id": "RL-032",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_us",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": "separation_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-033",
   "case_id": "RL-033",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_uk",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-034",
   "case_id": "RL-034",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_uk",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-035",
   "case_id": "RL-035",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "msa_dpa",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "msa",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-036",
   "case_id": "RL-036",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "sports_sponsorship",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-037",
   "case_id": "RL-037",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "sports_sponsorship",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-038",
   "case_id": "RL-038",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "athlete",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-039",
   "case_id": "RL-039",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "athlete",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": "athlete_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-040",
   "case_id": "RL-040",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "ticketing",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-041",
   "case_id": "RL-041",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "event_rights",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-042",
   "case_id": "RL-042",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "sports_sponsorship",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-043",
   "case_id": "RL-043",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "demand_letter",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": "demand_letter",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-044",
   "case_id": "RL-044",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "lease",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-045",
   "case_id": "RL-045",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "family",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-046",
   "case_id": "RL-046",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "family",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-047",
   "case_id": "RL-047",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_us",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-048",
   "case_id": "RL-048",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "smb_contract",
   "jurisdiction": "US-AZ",
   "customer": null,
   "doc_type": "services_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-049",
   "case_id": "RL-049",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "signing_recipient",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-050",
   "case_id": "RL-050",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "signing_recipient",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-051",
   "case_id": "RL-051",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "signing_recipient",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-052",
   "case_id": "RL-052",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "uae_cross_border",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-053",
   "case_id": "RL-053",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "uae_cross_border",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-054",
   "case_id": "RL-054",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "uae_cross_border",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-055",
   "case_id": "RL-055",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_uk",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-056",
   "case_id": "RL-056",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_uk",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": "settlement_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-057",
   "case_id": "RL-057",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "uae_cross_border",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "services_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-058",
   "case_id": "RL-058",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "eu_cross_border",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-059",
   "case_id": "RL-059",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-060",
   "case_id": "RL-060",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-061",
   "case_id": "RL-061",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-062",
   "case_id": "RL-062",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-063",
   "case_id": "RL-063",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-064",
   "case_id": "RL-064",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-065",
   "case_id": "RL-065",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-066",
   "case_id": "RL-066",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-067",
   "case_id": "RL-067",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-068",
   "case_id": "RL-068",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "adversarial",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-069",
   "case_id": "RL-069",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "nda",
   "jurisdiction": "US-NY",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-070",
   "case_id": "RL-070",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-071",
   "case_id": "RL-071",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "contractor_ip",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-072",
   "case_id": "RL-072",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "employment_us",
   "jurisdiction": "US-TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-073",
   "case_id": "RL-073",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "privacy",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "privacy_policy",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-074",
   "case_id": "RL-074",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "delaware_ops",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-075",
   "case_id": "RL-075",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "vc_term_sheet",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-076",
   "case_id": "RL-076",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "sports_sponsorship",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": "sponsorship_agreement",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-077",
   "case_id": "RL-077",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "smb_contract",
   "jurisdiction": "US-OR",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-078",
   "case_id": "RL-078",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "msa_dpa",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-079",
   "case_id": "RL-079",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "board_consent",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "stockholder_consent",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-080",
   "case_id": "RL-080",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "privacy",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-081",
   "case_id": "RL-081",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "sports_sponsorship",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": "board_resolution",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-082",
   "case_id": "RL-082",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "smb_contract",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "non_renewal_notice",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-083",
   "case_id": "RL-083",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-084",
   "case_id": "RL-084",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-085",
   "case_id": "RL-085",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-086",
   "case_id": "RL-086",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-087",
   "case_id": "RL-087",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-088",
   "case_id": "RL-088",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "demand_letter",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-089",
   "case_id": "RL-089",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-090",
   "case_id": "RL-090",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-091",
   "case_id": "RL-091",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-092",
   "case_id": "RL-092",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-093",
   "case_id": "RL-093",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "board_consent",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-094",
   "case_id": "RL-094",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": "ticket_terms",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-095",
   "case_id": "RL-095",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-096",
   "case_id": "RL-096",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-097",
   "case_id": "RL-097",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "sponsorship_term_sheet",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-098",
   "case_id": "RL-098",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": "UK-EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-099",
   "case_id": "RL-099",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-100",
   "case_id": "RL-100",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": "US-CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-101",
   "case_id": "RL-101",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "safe",
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "case:RL-102",
   "case_id": "RL-102",
   "kind": "routing_case",
   "title": null,
   "source": "realistic",
   "use_case": "followup",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "CASE (not yet run)",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/routing/cases.jsonl",
   "notes": null,
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-001",
   "case_id": "RC-001",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.03,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing ['scope_of_waiver_limited']; must_not ['\\\\bMy company\\\\b', '(?im)^#*\\\\s*(Simple Agreement for Future Equity|SAFE)\\\\s*$']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-002",
   "case_id": "RC-002",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0381,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing ['rest_of_agreement_unchanged']; must_not ['\\\\bMy company\\\\b', '2,000,000', 'Purchase Amount\\\\W{0,6}(\\\\(|means|of|:|is)?\\\\W{0,6}(US)?\\\\$\\\\s?\\\\d|\\\\$\\\\s?[\\\\d,.]+\\\\s?(M|million|K)?\\\\s?\\\\(the \\\\W?Purchase Amount']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-003",
   "case_id": "RC-003",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0353,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing ['references_side_letter', 'multiple_investors']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-004",
   "case_id": "RC-004",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0072,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing ['references_prior_instrument', 'rest_of_agreement_unchanged']; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-005",
   "case_id": "RC-005",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0349,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing ['scope_of_waiver_limited']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-006",
   "case_id": "RC-006",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0321,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['identifies_mfn_clause', 'references_side_letter']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-007",
   "case_id": "RC-007",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0359,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-008",
   "case_id": "RC-008",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0311,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['eighteen_months']; must_not ['\\\\bMy company\\\\b', '18,000,000']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-009",
   "case_id": "RC-009",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0201,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_mutual; missing ['governing_law_delaware']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-010",
   "case_id": "RC-010",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0207,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_mutual; missing ['one_way_roles', 'three_year_term']; must_not ['\\\\bMy company\\\\b', '3,000,000']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-011",
   "case_id": "RC-011",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0206,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_one_way; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-012",
   "case_id": "RC-012",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0206,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_one_way; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-013",
   "case_id": "RC-013",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "advisor",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0279,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['services_scope', 'equity_grant', 'vesting_cliff', 'independent_status']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-014",
   "case_id": "RC-014",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0348,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing ['no_discount']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-015",
   "case_id": "RC-015",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0339,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected safe; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-016",
   "case_id": "RC-016",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0207,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-017",
   "case_id": "RC-017",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0202,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-018",
   "case_id": "RC-018",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0204,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_mutual; missing ['one_way_roles']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-019",
   "case_id": "RC-019",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0258,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['independent_status', 'fee_aed', 'six_months', 'termination']; must_not ['\\\\bMy company\\\\b', '\\\\$\\\\s?\\\\d']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-020",
   "case_id": "RC-020",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0313,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['scope_of_services', 'fee_aed', 'milestones', 'ip_assignment']; must_not ['\\\\bMy company\\\\b', '\\\\$\\\\s?\\\\d']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-021",
   "case_id": "RC-021",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0269,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['scope_of_services', 'fee_aed', 'twelve_months', 'limitation_of_liability', 'termination']; must_not ['\\\\bMy company\\\\b', '\\\\$\\\\s?\\\\d']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-022",
   "case_id": "RC-022",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0309,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['independent_status', 'hourly_rate_aed', 'hours_cap', 'termination']; must_not ['\\\\bMy company\\\\b', '\\\\$\\\\s?\\\\d']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-023",
   "case_id": "RC-023",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0211,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-024",
   "case_id": "RC-024",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0249,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1.json",
   "notes": "detected term_sheet; missing ['independent_status', 'fee_aed', 'three_months', 'ip_assignment', 'termination']; must_not ['\\\\bMy company\\\\b', '\\\\$\\\\s?\\\\d']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-025",
   "case_id": "RC-025",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.032,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1-qa.json",
   "notes": "detected safe; missing ['references_prior_instrument', 'scope_new_round', 'rest_of_agreement_unchanged']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-026",
   "case_id": "RC-026",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0389,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1-qa.json",
   "notes": "detected safe; missing ['scope_new_round', 'both_investors']; must_not ['\\\\bMy company\\\\b']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_pascal-1:RC-027",
   "case_id": "RC-027",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "production prompt, pascal-1",
   "run": "ROUTE-C1 main_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0296,
   "latency_s": null,
   "link": "evals/routing/runs/main-pascal-1-qa.json",
   "notes": "detected term_sheet; missing ['advisor_title', 'signs_personally']; must_not ['\\\\bMy company\\\\b', '\\\\b\\\\d+(\\\\.\\\\d+)?\\\\s?%\\\\s+(of the|of fully|fully)']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_astra:RC-001",
   "case_id": "RC-001",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "openai/gpt-6-astra",
   "prompt_version": "production prompt, gpt-6-astra",
   "run": "ROUTE-C1 main_astra",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.1595,
   "latency_s": null,
   "link": "evals/routing/runs/main-astra.json",
   "notes": "detected safe; missing ['rest_of_agreement_unchanged']; must_not ['\\\\bMy company\\\\b', '(?im)^#*\\\\s*(Simple Agreement for Future Equity|SAFE)\\\\s*$']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_astra:RC-002",
   "case_id": "RC-002",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "openai/gpt-6-astra",
   "prompt_version": "production prompt, gpt-6-astra",
   "run": "ROUTE-C1 main_astra",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.1457,
   "latency_s": null,
   "link": "evals/routing/runs/main-astra.json",
   "notes": "detected safe; missing ['waiver_operative', 'rest_of_agreement_unchanged']; must_not ['\\\\bMy company\\\\b', '2,000,000']; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "main_astra:RC-017",
   "case_id": "RC-017",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "openai/gpt-6-astra",
   "prompt_version": "production prompt, gpt-6-astra",
   "run": "ROUTE-C1 main_astra",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.1182,
   "latency_s": null,
   "link": "evals/routing/runs/main-astra.json",
   "notes": "detected nda_mutual; missing []; must_not ['\\\\bMy company\\\\b']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "main_astra:RC-019",
   "case_id": "RC-019",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "openai/gpt-6-astra",
   "prompt_version": "production prompt, gpt-6-astra",
   "run": "ROUTE-C1 main_astra",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.1348,
   "latency_s": null,
   "link": "evals/routing/runs/main-astra.json",
   "notes": "detected term_sheet; missing ['independent_status', 'fee_aed', 'six_months', 'termination']; must_not ['\\\\bMy company\\\\b', '\\\\$\\\\s?\\\\d']; jurisdiction miss ['governing_law', 'courts']",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-001",
   "case_id": "RC-001",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0131,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-002",
   "case_id": "RC-002",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.013,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-013",
   "case_id": "RC-013",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "advisor",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0364,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected advisor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-016",
   "case_id": "RC-016",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0261,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-017",
   "case_id": "RC-017",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0266,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-018",
   "case_id": "RC-018",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0231,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected nda_one_way; missing []; must_not []; jurisdiction miss ['governing_law']",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-019",
   "case_id": "RC-019",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.034,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected contractor_agreement; missing ['six_months']; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-020",
   "case_id": "RC-020",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0471,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected services_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-021",
   "case_id": "RC-021",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0447,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected services_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "eng_pascal-1:RC-026",
   "case_id": "RC-026",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "request passed through, pascal-1",
   "run": "ROUTE-C1 eng_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0156,
   "latency_s": null,
   "link": "evals/routing/runs/eng-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-001",
   "case_id": "RC-001",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0115,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-002",
   "case_id": "RC-002",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0123,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-003",
   "case_id": "RC-003",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0152,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-004",
   "case_id": "RC-004",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0117,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-005",
   "case_id": "RC-005",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0109,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-006",
   "case_id": "RC-006",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0113,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-007",
   "case_id": "RC-007",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0233,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected side_letter; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-008",
   "case_id": "RC-008",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "side_letter",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "side_letter",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0191,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected side_letter; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-009",
   "case_id": "RC-009",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0196,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-010",
   "case_id": "RC-010",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.019,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_one_way; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-011",
   "case_id": "RC-011",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0183,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-012",
   "case_id": "RC-012",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "nda",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0213,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-013",
   "case_id": "RC-013",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "advisor",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "false",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0315,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected advisor_agreement; missing ['vesting_cliff']; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-014",
   "case_id": "RC-014",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0428,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected safe; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-015",
   "case_id": "RC-015",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "safe",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "safe",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0405,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected safe; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-016",
   "case_id": "RC-016",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0223,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-017",
   "case_id": "RC-017",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0232,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-018",
   "case_id": "RC-018",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_one_way",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0206,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_one_way; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-019",
   "case_id": "RC-019",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0321,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected contractor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-020",
   "case_id": "RC-020",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0392,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected services_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-021",
   "case_id": "RC-021",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "services_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0385,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected services_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-022",
   "case_id": "RC-022",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "DIFC",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0304,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected contractor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-023",
   "case_id": "RC-023",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_nda",
   "jurisdiction": "ADGM",
   "customer": null,
   "doc_type": "nda_mutual",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0199,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected nda_mutual; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-024",
   "case_id": "RC-024",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0309,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1.json",
   "notes": "detected contractor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-025",
   "case_id": "RC-025",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0131,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1-qa.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-026",
   "case_id": "RC-026",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.014,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1-qa.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_pascal-1:RC-027",
   "case_id": "RC-027",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v1, pascal-1",
   "run": "ROUTE-C1 fix_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0245,
   "latency_s": null,
   "link": "evals/routing/runs/fix-pascal-1-qa.json",
   "notes": "detected advisor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_v2_pascal-1:RC-001",
   "case_id": "RC-001",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "mfn_waiver",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "mfn_waiver",
   "model": "pascal-1",
   "prompt_version": "proposed fix v2, pascal-1",
   "run": "ROUTE-C1 fix_v2_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0111,
   "latency_s": null,
   "link": "evals/routing/runs/fix-v2-pascal-1.json",
   "notes": "detected mfn_waiver; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_v2_pascal-1:RC-013",
   "case_id": "RC-013",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "advisor",
   "jurisdiction": "US-DE",
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v2, pascal-1",
   "run": "ROUTE-C1 fix_v2_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0322,
   "latency_s": null,
   "link": "evals/routing/runs/fix-v2-pascal-1.json",
   "notes": "detected advisor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_v2_pascal-1:RC-019",
   "case_id": "RC-019",
   "kind": "drafting_case",
   "title": null,
   "source": "hill-climber",
   "use_case": "uae_services",
   "jurisdiction": "AE-ONSHORE",
   "customer": null,
   "doc_type": "contractor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v2, pascal-1",
   "run": "ROUTE-C1 fix_v2_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0332,
   "latency_s": null,
   "link": "evals/routing/runs/fix-v2-pascal-1.json",
   "notes": "detected contractor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "fix_v2_pascal-1:RC-027",
   "case_id": "RC-027",
   "kind": "drafting_case",
   "title": null,
   "source": "qa-pod",
   "use_case": "advisor",
   "jurisdiction": null,
   "customer": null,
   "doc_type": "advisor_agreement",
   "model": "pascal-1",
   "prompt_version": "proposed fix v2, pascal-1",
   "run": "ROUTE-C1 fix_v2_pascal-1",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "draft_pass",
   "metric_value": "true",
   "basis": "measured",
   "label": "MEASURED",
   "judge": "deterministic + Pascal-1 (invented facts)",
   "run_date": "2026-10-31",
   "cost_usd": 0.0285,
   "latency_s": null,
   "link": "evals/routing/runs/fix-v2-pascal-1.json",
   "notes": "detected advisor_agreement; missing []; must_not []; jurisdiction miss []",
   "segment": null
  },
  {
   "row_id": "customer:SEGA-1",
   "case_id": "SEGA-1",
   "kind": "customer_task",
   "title": "Regulated vendor data terms task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "customer:SEGA-2",
   "case_id": "SEGA-2",
   "kind": "customer_task",
   "title": "Regulated vendor data terms task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "customer:SEGA-3",
   "case_id": "SEGA-3",
   "kind": "customer_task",
   "title": "Regulated vendor data terms task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "customer:SEGA-4",
   "case_id": "SEGA-4",
   "kind": "customer_task",
   "title": "Regulated vendor data terms task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "customer:SEGA-5",
   "case_id": "SEGA-5",
   "kind": "customer_task",
   "title": "Regulated vendor data terms task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "customer:SEGA-6",
   "case_id": "SEGA-6",
   "kind": "customer_task",
   "title": "Regulated vendor data terms task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "customer:SEGB-1",
   "case_id": "SEGB-1",
   "kind": "customer_task",
   "title": "Sports sponsorship task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "customer:SEGB-2",
   "case_id": "SEGB-2",
   "kind": "customer_task",
   "title": "Sports sponsorship task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "customer:SEGB-3",
   "case_id": "SEGB-3",
   "kind": "customer_task",
   "title": "Sports sponsorship task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "customer:SEGB-4",
   "case_id": "SEGB-4",
   "kind": "customer_task",
   "title": "Sports sponsorship task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "customer:SEGB-5",
   "case_id": "SEGB-5",
   "kind": "customer_task",
   "title": "Sports sponsorship task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "customer:SEGB-6",
   "case_id": "SEGB-6",
   "kind": "customer_task",
   "title": "Sports sponsorship task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "customer:SEGC-1",
   "case_id": "SEGC-1",
   "kind": "customer_task",
   "title": "Investor outreach compliance task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Investor outreach compliance (customer-derived)"
  },
  {
   "row_id": "customer:SEGC-2",
   "case_id": "SEGC-2",
   "kind": "customer_task",
   "title": "Investor outreach compliance task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Investor outreach compliance (customer-derived)"
  },
  {
   "row_id": "customer:SEGC-3",
   "case_id": "SEGC-3",
   "kind": "customer_task",
   "title": "Investor outreach compliance task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Investor outreach compliance (customer-derived)"
  },
  {
   "row_id": "customer:SEGC-4",
   "case_id": "SEGC-4",
   "kind": "customer_task",
   "title": "Investor outreach compliance task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Investor outreach compliance (customer-derived)"
  },
  {
   "row_id": "customer:SEGC-5",
   "case_id": "SEGC-5",
   "kind": "customer_task",
   "title": "Investor outreach compliance task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Investor outreach compliance (customer-derived)"
  },
  {
   "row_id": "customer:SEGC-6",
   "case_id": "SEGC-6",
   "kind": "customer_task",
   "title": "Investor outreach compliance task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Investor outreach compliance (customer-derived)"
  },
  {
   "row_id": "customer:SEGD-1",
   "case_id": "SEGD-1",
   "kind": "customer_task",
   "title": "SaaS services agreements task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "SaaS services agreements (customer-derived)"
  },
  {
   "row_id": "customer:SEGD-2",
   "case_id": "SEGD-2",
   "kind": "customer_task",
   "title": "SaaS services agreements task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "SaaS services agreements (customer-derived)"
  },
  {
   "row_id": "customer:SEGD-3",
   "case_id": "SEGD-3",
   "kind": "customer_task",
   "title": "SaaS services agreements task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "SaaS services agreements (customer-derived)"
  },
  {
   "row_id": "customer:SEGD-4",
   "case_id": "SEGD-4",
   "kind": "customer_task",
   "title": "SaaS services agreements task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "SaaS services agreements (customer-derived)"
  },
  {
   "row_id": "customer:SEGD-5",
   "case_id": "SEGD-5",
   "kind": "customer_task",
   "title": "SaaS services agreements task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "SaaS services agreements (customer-derived)"
  },
  {
   "row_id": "customer:SEGE-1",
   "case_id": "SEGE-1",
   "kind": "customer_task",
   "title": "Overnight MSA review task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Overnight MSA review (customer-derived)"
  },
  {
   "row_id": "customer:SEGE-2",
   "case_id": "SEGE-2",
   "kind": "customer_task",
   "title": "Overnight MSA review task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Overnight MSA review (customer-derived)"
  },
  {
   "row_id": "customer:SEGE-3",
   "case_id": "SEGE-3",
   "kind": "customer_task",
   "title": "Overnight MSA review task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Overnight MSA review (customer-derived)"
  },
  {
   "row_id": "customer:SEGE-4",
   "case_id": "SEGE-4",
   "kind": "customer_task",
   "title": "Overnight MSA review task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Overnight MSA review (customer-derived)"
  },
  {
   "row_id": "customer:SEGE-5",
   "case_id": "SEGE-5",
   "kind": "customer_task",
   "title": "Overnight MSA review task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Overnight MSA review (customer-derived)"
  },
  {
   "row_id": "customer:SEGF-1",
   "case_id": "SEGF-1",
   "kind": "customer_task",
   "title": "Health data company set-up task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Health data company set-up (customer-derived)"
  },
  {
   "row_id": "customer:SEGF-2",
   "case_id": "SEGF-2",
   "kind": "customer_task",
   "title": "Health data company set-up task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Health data company set-up (customer-derived)"
  },
  {
   "row_id": "customer:SEGF-3",
   "case_id": "SEGF-3",
   "kind": "customer_task",
   "title": "Health data company set-up task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Health data company set-up (customer-derived)"
  },
  {
   "row_id": "customer:SEGF-4",
   "case_id": "SEGF-4",
   "kind": "customer_task",
   "title": "Health data company set-up task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Health data company set-up (customer-derived)"
  },
  {
   "row_id": "customer:SEGF-5",
   "case_id": "SEGF-5",
   "kind": "customer_task",
   "title": "Health data company set-up task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Health data company set-up (customer-derived)"
  },
  {
   "row_id": "customer:SEGG-1",
   "case_id": "SEGG-1",
   "kind": "customer_task",
   "title": "Venture financing documents task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Venture financing documents (customer-derived)"
  },
  {
   "row_id": "customer:SEGG-2",
   "case_id": "SEGG-2",
   "kind": "customer_task",
   "title": "Venture financing documents task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Venture financing documents (customer-derived)"
  },
  {
   "row_id": "customer:SEGG-3",
   "case_id": "SEGG-3",
   "kind": "customer_task",
   "title": "Venture financing documents task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Venture financing documents (customer-derived)"
  },
  {
   "row_id": "customer:SEGG-4",
   "case_id": "SEGG-4",
   "kind": "customer_task",
   "title": "Venture financing documents task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Venture financing documents (customer-derived)"
  },
  {
   "row_id": "customer:SEGG-5",
   "case_id": "SEGG-5",
   "kind": "customer_task",
   "title": "Venture financing documents task",
   "source": "customer-derived, anonymised",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "TARGET",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Target trace only; not yet run.",
   "segment": "Venture financing documents (customer-derived)"
  },
  {
   "row_id": "citation:0",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(2) permits general written authorisation; 28(3)(a) and (h) address instructions and audits.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:1",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=no; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(2) supports the proposition, but it has no further numbered subsection (2).",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:2",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3) exists; the repeated (3) is erroneous. Legal compulsion is excepted, and notice concerns the processor’s opinion.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:3",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28(3)(h)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "It requires audit and inspection provisions but does not mandate unrestricted, any-time on-site access.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:4",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(4) makes the initial processor fully liable to the controller when the sub-processor fails its data protection obligations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:5",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 33(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(1) applies where notification is required; 72 hours is qualified by feasibility, and delays require reasons.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:6",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 33(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) states this requirement; the second “(2)” is redundant.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:7",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 33",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(3) specifies the minimum information in a controller’s breach notification.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:8",
   "case_id": "03",
   "kind": "citation",
   "title": "Chapter V",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=unsure; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The act is not identified. GDPR Chapter V governs international transfers, including adequacy decisions, safeguards, and derogations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:9",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 4(12)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/4",
   "notes": "Article 4(12) defines a personal data breach; the second “(12)” is redundant.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:10",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 46",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/46",
   "notes": "Article 46 lists appropriate safeguards, including standard data protection clauses, for third-country transfers.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:11",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 82(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82(1) provides compensation for infringement-related damage; it does not expressly address contractual liability caps.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:12",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 82(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82(4) imposes liability for the entire damage when multiple controllers or processors are responsible.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:13",
   "case_id": "03",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The GDPR establishes controller and processor duties; the DPA, not the GDPR, defines its contractual term.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:14",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision adopts standard contractual clauses for transfers of personal data to third countries under the GDPR.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:15",
   "case_id": "03",
   "kind": "citation",
   "title": "Clause 9, Option 2 of the SCCs",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Clause 9, Option 2 requires advance written notice of intended sub-processor additions or replacements.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:16",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2023/1795",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision covers transfers to US organisations included on the Data Privacy Framework List.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:17",
   "case_id": "03",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 2(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_2.html",
   "notes": "Article 2(1) lists covered entities; paragraph 2 defines financial entities, and paragraphs 3 to 4 provide exclusions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:18",
   "case_id": "03",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 30",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30 sets contractual requirements for in-scope financial entities, but also addresses ICT providers and their contractual obligations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:19",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR 255.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.5",
   "notes": "Section 255.5(a) requires clear and conspicuous disclosure when the material connection is not reasonably expected by the audience.",
   "segment": null
  },
  {
   "row_id": "citation:20",
   "case_id": "06",
   "kind": "citation",
   "title": "CAP Code Section 2",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "CAP Code rule 2.1 states that marketing communications must be obviously identifiable as such.",
   "segment": null
  },
  {
   "row_id": "citation:21",
   "case_id": "06",
   "kind": "citation",
   "title": "FTC Endorsement Guides, 16 CFR Part 255 (2023 update)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The Guides address endorsement advertising, but disclosure is required for material connections not reasonably expected by consumers, not every endorsement.",
   "segment": null
  },
  {
   "row_id": "citation:22",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 28",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 28 requires financial entities, including banks, to manage ICT third-party risk and address contractual arrangements.",
   "segment": null
  },
  {
   "row_id": "citation:23",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Arts. 28 and 30(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(2) specifies minimum contractual elements; Article 28 sets related ICT third-party risk management principles.",
   "segment": null
  },
  {
   "row_id": "citation:24",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Applies to assistance for ICT incidents related to the ICT service provided to the financial entity.",
   "segment": null
  },
  {
   "row_id": "citation:25",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 30(3)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(3) requires additional contractual terms; the repeated (3) is redundant.",
   "segment": null
  },
  {
   "row_id": "citation:26",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 19",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_19.html",
   "notes": "Article 19(1) places the duty on financial entities; paragraph 5 permits outsourcing while preserving their responsibility.",
   "segment": null
  },
  {
   "row_id": "citation:27",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301, Art. 5(1)(a)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 5(1)(a) sets the four-hour and 24-hour initial-notification deadlines.",
   "segment": null
  },
  {
   "row_id": "citation:28",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "PRA SS2/21",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SS2/21 sets expectations for material outsourcing but imposes no blanket two-hour notice or unlimited-liability requirement.",
   "segment": null
  },
  {
   "row_id": "citation:29",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 33(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The 72-hour period runs from awareness, where feasible; notification is unnecessary if the breach is unlikely to pose risks.",
   "segment": null
  },
  {
   "row_id": "citation:30",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 33(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 33(2) requires notification without undue delay after the processor becomes aware of the breach.",
   "segment": null
  },
  {
   "row_id": "citation:31",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 4(12)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 4(12) defines ‘personal data breach’ under the GDPR.",
   "segment": null
  },
  {
   "row_id": "citation:32",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 82",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 82 requires material or non-material damage; it does not expressly address contractual liability caps.",
   "segment": null
  },
  {
   "row_id": "citation:33",
   "case_id": "01",
   "kind": "citation",
   "title": "DGCL s.152",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 152 permits board authorization and determination of consideration, subject to delegation and statutory limitations.",
   "segment": null
  },
  {
   "row_id": "citation:34",
   "case_id": "01",
   "kind": "citation",
   "title": "DGCL s.141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Written consent requires all board members’ consent and may also be restricted by the bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:35",
   "case_id": "01",
   "kind": "citation",
   "title": "YC post-money Safe, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) permits majority-in-interest amendments subject to the stated safeguards.",
   "segment": null
  },
  {
   "row_id": "citation:36",
   "case_id": "CG2",
   "kind": "citation",
   "title": "YC post-money SAFE, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) governs amendments to the SAFE; it does not expressly address separate side letters.",
   "segment": null
  },
  {
   "row_id": "citation:37",
   "case_id": "18",
   "kind": "citation",
   "title": "California Civil Code section 1950.5(h)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "The provision requires an itemized statement and return of any remaining security deposit within 21 calendar days after vacancy.",
   "segment": null
  },
  {
   "row_id": "citation:38",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code section 1950.5(m)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "The bad-faith damages provision is in subdivision (l), not subdivision (m).",
   "segment": null
  },
  {
   "row_id": "citation:39",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code s.1950.5(o)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Small claims jurisdiction depends on the amount claimed being within applicable limits.",
   "segment": null
  },
  {
   "row_id": "citation:40",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.221",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.221",
   "notes": "Section 116.221 sets the $12,500 limit for actions brought by natural persons, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:41",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.530(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "Section 116.530(a) generally prohibits attorney participation, subject to exceptions specified in the section.",
   "segment": null
  },
  {
   "row_id": "citation:42",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.940",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.940",
   "notes": "Section 116.260 imposes the service requirement; 116.940 governs it, permits court administration, and provides limited exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:43",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.230",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.230",
   "notes": "Fees vary by claim size, with a separate fee for frequent filers.",
   "segment": null
  },
  {
   "row_id": "citation:44",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.340",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.340",
   "notes": "Subdivision (b) sets service deadlines of 15 or 20 days before the hearing, depending on residence.",
   "segment": null
  },
  {
   "row_id": "citation:45",
   "case_id": "18",
   "kind": "citation",
   "title": "Judicial Council form SC-100",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "California Judicial Council form SC-100 is used to file a plaintiff’s small claims case.",
   "segment": null
  },
  {
   "row_id": "citation:46",
   "case_id": "18",
   "kind": "citation",
   "title": "FW-001",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "California Judicial Council form FW-001 is the Request to Waive Court Fees.",
   "segment": null
  },
  {
   "row_id": "citation:47",
   "case_id": "18",
   "kind": "citation",
   "title": "SC-104",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "California form SC-104 is the Proof of Service form for small claims cases.",
   "segment": null
  },
  {
   "row_id": "citation:48",
   "case_id": "CG1",
   "kind": "citation",
   "title": "YC post-money SAFE, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The solicitation, equal-treatment and Purchase Amount conditions apply to the majority-consent route.",
   "segment": null
  },
  {
   "row_id": "citation:49",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. § 1839",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "Section 1839(3) defines trade secret for the chapter, including the DTSA.",
   "segment": null
  },
  {
   "row_id": "citation:50",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. § 1839(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "Section 1839(3) defines trade secret; the repeated (3) in the citation is redundant.",
   "segment": null
  },
  {
   "row_id": "citation:51",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Section 141(f) of the General Corporation Law of the State of Delaware",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) requires unanimous consent filed with the minutes, subject to restrictions in the certificate of incorporation or bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:52",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) requires unanimous written or electronic consent filed with the board’s minutes.",
   "segment": null
  },
  {
   "row_id": "citation:53",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §141(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Management may be by or under the board’s direction, subject to statutory and certificate exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:54",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §152",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 152(a) supports the proposition; subsection (b) permits delegation subject to specified limits.",
   "segment": null
  },
  {
   "row_id": "citation:55",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §242(b)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Board approval is generally required, but some amendments do not require a stockholder vote.",
   "segment": null
  },
  {
   "row_id": "citation:56",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The EU AI Act uses a risk-based framework and imposes obligations on providers and deployers.",
   "segment": null
  },
  {
   "row_id": "citation:57",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/6/",
   "notes": "Article 6 sets high-risk classification rules; the agents’ classification depends on their intended uses and specific features.",
   "segment": null
  },
  {
   "row_id": "citation:58",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Annex III",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III includes recruitment and certain emotion-recognition uses, but not sales outreach or meeting booking as such.",
   "segment": null
  },
  {
   "row_id": "citation:59",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Annex I",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex I lists the legislation; Article 6(1) sets additional conditions for high-risk classification.",
   "segment": null
  },
  {
   "row_id": "citation:60",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 2(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/2/",
   "notes": "Article 2(1)(a) and (c) support the stated territorial scope.",
   "segment": null
  },
  {
   "row_id": "citation:61",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 3(3), Art. 3(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/3/",
   "notes": "The provisions define both roles; Client M’s and the customer’s classifications depend on facts not provided.",
   "segment": null
  },
  {
   "row_id": "citation:62",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 3(60)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/3/",
   "notes": "The definition also requires content to falsely appear authentic or truthful and includes AI-manipulated content.",
   "segment": null
  },
  {
   "row_id": "citation:63",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/4/",
   "notes": "Article 4 requires providers and deployers to take measures, to their best extent, not guarantee staff knowledge.",
   "segment": null
  },
  {
   "row_id": "citation:64",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Article 5 prohibits specified AI practices; it applies from 2 February 2025 under Article 113(a).",
   "segment": null
  },
  {
   "row_id": "citation:65",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 22",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/22/",
   "notes": "Article 22(1) supports the requirement before Union market availability; applicability here cannot be determined without the facts.",
   "segment": null
  },
  {
   "row_id": "citation:66",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 25(1)(c)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/25/",
   "notes": "Article 25(1)(c) treats the deployer as the provider when the changed purpose makes the system high-risk.",
   "segment": null
  },
  {
   "row_id": "citation:67",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 50(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(5) specifies timing; Article 50(1) includes exceptions, including where AI interaction is obvious.",
   "segment": null
  },
  {
   "row_id": "citation:68",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 50(2), (4), (5)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Text disclosure concerns matters of public interest and is subject to exceptions, including human editorial review and responsibility.",
   "segment": null
  },
  {
   "row_id": "citation:69",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 99",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/99/",
   "notes": "Article 99 sets administrative fine ceilings for specified infringements, with special rules for SMEs.",
   "segment": null
  },
  {
   "row_id": "citation:70",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/113/",
   "notes": "Article 4 and Article 50 dates are correct, but the supplied text lists exceptions for Chapter II.",
   "segment": null
  },
  {
   "row_id": "citation:71",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Annex III point 1(c)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III, point 1(c), lists AI systems intended for emotion recognition.",
   "segment": null
  },
  {
   "row_id": "citation:72",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Annex III point 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III point 4 covers these uses, subject to Article 6’s high-risk classification conditions.",
   "segment": null
  },
  {
   "row_id": "citation:73",
   "case_id": "13",
   "kind": "citation",
   "title": "Directive 2002/58/EC, Art. 13(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eudr/2002/58/article/13",
   "notes": "Consent is required, but the existing-customer email exception is specified in Article 13(2), not an open-ended national exception.",
   "segment": null
  },
  {
   "row_id": "citation:74",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 30",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "x-ai/grok-4.7",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30 sets ICT contractual requirements; it does not address AI Act classification.",
   "segment": null
  },
  {
   "row_id": "citation:75",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28 GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(h) requires compliance information and cooperation with audits under the processor contract.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:76",
   "case_id": "03",
   "kind": "citation",
   "title": "Chapter V GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Chapter V governs transfers of personal data to third countries or international organisations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:77",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 45 GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/45",
   "notes": "Article 45 permits EU GDPR transfers based on Commission adequacy decisions; the source shows its omission from UK GDPR.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:78",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 46 GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/46",
   "notes": "Article 46(2) lists standard data protection clauses as appropriate safeguards; the supplied text is the UK version.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:79",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Article 28(3)(a), final paragraph",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The warning duty appears in Article 28(3)’s final subparagraph, not (3)(a); legal requirements may override instructions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:80",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 28(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Notice of intended changes and opportunity to object apply specifically to general written authorisation.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:81",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 28(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(4) supports both obligations; the citation repeats the subsection number.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:82",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Article 28(3)(h)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The processor’s contract or other legal act must stipulate these information and audit obligations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:83",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 33(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Notification is not required if the breach is unlikely to risk individuals’ rights and freedoms; 72 hours runs from awareness.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:84",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 33(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) expressly states this duty; the citation repeats the paragraph number.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:85",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 82(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Processor liability is limited to breaches of processor-specific obligations or acting outside or contrary to lawful controller instructions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:86",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 82(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Full liability applies where parties are involved in the same processing and are responsible under Article 82(2) to (3).",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:87",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 82(5)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82(5) provides recourse after full compensation, proportionate to the other parties’ responsibility.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:88",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR, Articles 44 to 46",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Articles 44 to 46 address general transfer requirements, adequacy decisions, and appropriate safeguards, respectively.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:89",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914, Article 1 and Annex",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision provides standard contractual clauses in an annex with modules for different transfer relationships.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:90",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914, Modules Two and Three, Clause 14",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Clause 14 requires a documented assessment of destination-country laws and practices affecting compliance with the clauses.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:91",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2023/1795, Article 1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 1 finds adequate protection for transfers to organisations on the Data Privacy Framework List.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:92",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR §§ 255.1(d)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.1",
   "notes": "Honest opinions are addressed in subsection (a); subsection (d) addresses misleading endorsements and undisclosed unexpected material connections.",
   "segment": null
  },
  {
   "row_id": "citation:93",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR § 255.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.5",
   "notes": "Disclosure is required when the material connection is not reasonably expected by a significant minority of the audience.",
   "segment": null
  },
  {
   "row_id": "citation:94",
   "case_id": "06",
   "kind": "citation",
   "title": "UK CAP Code, rule 2.1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 2.1 states that marketing communications must be obviously identifiable as such.",
   "segment": null
  },
  {
   "row_id": "citation:95",
   "case_id": "06",
   "kind": "citation",
   "title": "UK CAP Code, rule 2.3",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 2.3 requires commercial intent to be clear where it is not obvious from the context.",
   "segment": null
  },
  {
   "row_id": "citation:96",
   "case_id": "06",
   "kind": "citation",
   "title": "UK CAP Code, rule 3.7",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 3.7 requires documentary evidence for objectively substantiable claims before publication.",
   "segment": null
  },
  {
   "row_id": "citation:97",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 28(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 28(1) requires third-party risk management; contractual protections are addressed elsewhere, including Articles 28(7) and 30.",
   "segment": null
  },
  {
   "row_id": "citation:98",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "The requirement applies to assistance with ICT incidents related to the services provided to the financial entity.",
   "segment": null
  },
  {
   "row_id": "citation:99",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Art. 30(3)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(3) requires additional contractual terms for ICT services supporting critical or important functions.",
   "segment": null
  },
  {
   "row_id": "citation:100",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301, Art. 5(1)(a)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=unsure; accurate=no",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The four-hour period runs from classification to initial notification; it is not a deadline for classification.",
   "segment": null
  },
  {
   "row_id": "citation:101",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, Art. 33(1) to (2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Notification is required only for breaches likely to risk rights and freedoms; 72 hours applies where feasible, after awareness.",
   "segment": null
  },
  {
   "row_id": "citation:102",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, Art. 82(2), (4) to (5)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 82(2), (4) and (5) address liability for damage, full compensation, and contribution between responsible controllers and processors.",
   "segment": null
  },
  {
   "row_id": "citation:103",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "PRA SS2/21, Chapters 6 to 10",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Chapters 6 to 10 address these outsourcing and third-party risk management expectations.",
   "segment": null
  },
  {
   "row_id": "citation:104",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "FCA Handbook, SYSC 8.1.6R",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The rule retains the firm’s regulatory responsibility when outsourcing critical or important operational functions, subject to its applicability.",
   "segment": null
  },
  {
   "row_id": "citation:105",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "FCA Handbook, SYSC 8.1.8R",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Sets out safeguards for outsourcing critical or important operational functions.",
   "segment": null
  },
  {
   "row_id": "citation:106",
   "case_id": "01",
   "kind": "citation",
   "title": "6 Del. C. § 12A-107",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 12A-107 recognizes electronic records, signatures, and contracts, subject to the Act’s applicability and exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:107",
   "case_id": "01",
   "kind": "citation",
   "title": "Y Combinator, U.S. post-money Safe, § 1(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 1(a) addresses conversion using the Safe Price; the cap and pricing formula appear elsewhere.",
   "segment": null
  },
  {
   "row_id": "citation:108",
   "case_id": "01",
   "kind": "citation",
   "title": "Y Combinator, U.S. post-money Safe, § 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) addresses amendments; equity-financing conversion is addressed in Section 1(a).",
   "segment": null
  },
  {
   "row_id": "citation:109",
   "case_id": "CG2",
   "kind": "citation",
   "title": "6 Del. C. §12A-107(a), (b), (d)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 12A-107 recognizes the legal effect of electronic records, contracts, and signatures.",
   "segment": null
  },
  {
   "row_id": "citation:110",
   "case_id": "18",
   "kind": "citation",
   "title": "California Civil Code section 1950.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Section 1950.5(g)(1) generally requires return of remaining security and an itemized deduction statement within 21 calendar days.",
   "segment": null
  },
  {
   "row_id": "citation:111",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code §1950.5(h)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Subdivision (h)(1) requires an itemized accounting and return of remaining security within 21 calendar days after the tenant vacates.",
   "segment": null
  },
  {
   "row_id": "citation:112",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code §1950.5(b)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Subdivision (b) lists permissible security deposit purposes, including unpaid rent, damage repairs, and cleaning. Repeated '(b)' appears typographical.",
   "segment": null
  },
  {
   "row_id": "citation:113",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code §1950.5(m)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "The supplied excerpt omits subdivision (m), preventing verification of its damages and burden provisions.",
   "segment": null
  },
  {
   "row_id": "citation:114",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code §1950.5(o)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Subdivision (o) permits actions in small claims court for actual or statutory damages within applicable jurisdictional limits.",
   "segment": null
  },
  {
   "row_id": "citation:115",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. §116.221",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.221",
   "notes": "Sets a $12,500 small claims demand limit for natural persons, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:116",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. §116.230(b)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.230",
   "notes": "Subdivision (b) sets small claims filing fees by demand amount; the repeated '(b)' is a typographical duplication.",
   "segment": null
  },
  {
   "row_id": "citation:117",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. §116.340",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.340",
   "notes": "Specifies small claims service methods, completion deadlines before hearing, and proof-of-service filing timing.",
   "segment": null
  },
  {
   "row_id": "citation:118",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. §116.530(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "Subdivision (a) restricts attorney participation in small claims actions, subject to exceptions. The repeated '(a)' is a typographical duplication.",
   "segment": null
  },
  {
   "row_id": "citation:119",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. §116.530(c)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "Subdivision (c)(1) permits attorney advice to small claims parties before or after commencement; citation redundantly repeats the subsection.",
   "segment": null
  },
  {
   "row_id": "citation:120",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. §116.940",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.940",
   "notes": "Section 116.940 governs small claims advisory services, including assistance, adviser qualifications, administration, and immunity.",
   "segment": null
  },
  {
   "row_id": "citation:121",
   "case_id": "CG1",
   "kind": "citation",
   "title": "YC post-money SAFE, §5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) conditions majority amendments on company consent and other safeguards; signed terms should be checked.",
   "segment": null
  },
  {
   "row_id": "citation:122",
   "case_id": "CG3",
   "kind": "citation",
   "title": "N.Y. General Obligations Law § 5-1103",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The writing must be signed by the party against whom enforcement is sought or its agent.",
   "segment": null
  },
  {
   "row_id": "citation:123",
   "case_id": "CG3",
   "kind": "citation",
   "title": "Ashland Management Inc. v. Janien, 82 N.Y.2d 395, 407 (1993)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The cited page sets out factors for determining whether information qualifies as a trade secret.",
   "segment": null
  },
  {
   "row_id": "citation:124",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Section 141(f) of the Delaware General Corporation Law",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Subject to restrictions in the certificate of incorporation or bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:125",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) permits unanimous written or electronic consent without a meeting, subject to corporate-document restrictions.",
   "segment": null
  },
  {
   "row_id": "citation:126",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §116",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 116 permits electronic documentation and signatures for corporate acts, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:127",
   "case_id": "CG4",
   "kind": "citation",
   "title": "17 C.F.R. §230.506(b)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.506",
   "notes": "Rule 506(b) provides a conditional private-offering exemption; the duplicated (b) appears to be a citation typo.",
   "segment": null
  },
  {
   "row_id": "citation:128",
   "case_id": "CG4",
   "kind": "citation",
   "title": "17 C.F.R. §230.503(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.503",
   "notes": "Section 503(a)(1) provides the deadline, extended to the next business day for weekends or holidays.",
   "segment": null
  },
  {
   "row_id": "citation:129",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Article 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/6/",
   "notes": "Article 6 sets high-risk classification rules, but the limited facts do not establish classification of Client M’s specific uses.",
   "segment": null
  },
  {
   "row_id": "citation:130",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Annex III",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Insurance risk assessment is limited to life and health insurance; sales outreach is not expressly listed.",
   "segment": null
  },
  {
   "row_id": "citation:131",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Article 50(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "The obligation applies to providers; technical feasibility qualifies the required effectiveness, interoperability, robustness and reliability of their solutions.",
   "segment": null
  },
  {
   "row_id": "citation:132",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Article 50(1), (2), (4) and (5)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "The cited paragraphs address interaction disclosure, machine-readable marking, deepfakes and the timing of transparency information.",
   "segment": null
  },
  {
   "row_id": "citation:133",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Articles 2(1), 3(3) to (4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/2/",
   "notes": "Article 2(1) sets scope; Article 3(3) to (4) defines provider and deployer.",
   "segment": null
  },
  {
   "row_id": "citation:134",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Articles 4, 5(1)(a), (b), (f), and 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/4/",
   "notes": "Article 4 addresses AI literacy; Article 5 prohibits specified practices; Article 113 sets entry-into-force and application dates.",
   "segment": null
  },
  {
   "row_id": "citation:135",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Article 5(1)(a), (b), (f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Manipulation and exploitation require significant harm or its likelihood; workplace emotion inference has medical and safety exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:136",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Article 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/113/",
   "notes": "Article 113 sets entry into force and application dates, including staggered dates for specified provisions.",
   "segment": null
  },
  {
   "row_id": "citation:137",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Articles 43, 48 and 49",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/43/",
   "notes": "The articles address these requirements, but applicability cannot be determined without details of the sales uses.",
   "segment": null
  },
  {
   "row_id": "citation:138",
   "case_id": "13",
   "kind": "citation",
   "title": "Directive 2002/58/EC, Article 13(1) to (5)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eudr/2002/58/article/13",
   "notes": "The existing-customer exception applies to similar products or services and requires an opt-out at collection and each message.",
   "segment": null
  },
  {
   "row_id": "citation:139",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Article 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/6",
   "notes": "Article 6 includes consent and legitimate interests; marketing remains subject to applicable additional rules.",
   "segment": null
  },
  {
   "row_id": "citation:140",
   "case_id": "13",
   "kind": "citation",
   "title": "GDPR, Article 13",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/13",
   "notes": "Article 13 requires controllers to provide specified information when collecting personal data from the data subject, subject to exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:141",
   "case_id": "13",
   "kind": "citation",
   "title": "GDPR, Article 14(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/14",
   "notes": "First-contact timing applies when the data are used to communicate with the subject; other deadlines and exceptions also apply.",
   "segment": null
  },
  {
   "row_id": "citation:142",
   "case_id": "13",
   "kind": "citation",
   "title": "GDPR, Article 21(2) to (3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/21",
   "notes": "Processing must stop for direct marketing purposes, including related profiling.",
   "segment": null
  },
  {
   "row_id": "citation:143",
   "case_id": "13",
   "kind": "citation",
   "title": "GDPR, Article 28",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3) also permits another binding legal act and requires processing on documented instructions, not necessarily instructions in the contract.",
   "segment": null
  },
  {
   "row_id": "citation:144",
   "case_id": "13",
   "kind": "citation",
   "title": "GDPR, Article 35",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/35",
   "notes": "Article 35(1) requires the controller to assess high risks to natural persons’ rights and freedoms before processing.",
   "segment": null
  },
  {
   "row_id": "citation:145",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "DORA Articles 28 and 30",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 30 requires incident assistance generally and additional reporting terms for services supporting critical or important functions.",
   "segment": null
  },
  {
   "row_id": "citation:146",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "DORA Articles 28 and 30",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(2)(f) covers incident assistance; Article 30(3)(b) adds reporting obligations. Neither requires uncapped liability.",
   "segment": null
  },
  {
   "row_id": "citation:147",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "DORA Article 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Requires contractual assistance for ICT incidents related to the service, at no additional cost or a predetermined cost.",
   "segment": null
  },
  {
   "row_id": "citation:148",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Article 28(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 28(1) requires financial entities to manage ICT third-party risk within their ICT risk management framework.",
   "segment": null
  },
  {
   "row_id": "citation:149",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, Article 30(3)(b)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Contracts must include additional provider-to-financial-entity reporting obligations for ICT services supporting critical or important functions.",
   "segment": null
  },
  {
   "row_id": "citation:150",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Article 33",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Controller notification is required only for breaches likely to pose a risk, within 72 hours where feasible.",
   "segment": null
  },
  {
   "row_id": "citation:151",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, Article 33(1) to (2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "The 72-hour deadline applies where feasible and notification is not required if the breach is unlikely to pose a risk.",
   "segment": null
  },
  {
   "row_id": "citation:152",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301, Article 5(1)(a)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 5(1)(a) sets the deadline for initial notification of major ICT-related incidents.",
   "segment": null
  },
  {
   "row_id": "citation:153",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, Article 82(2) to (5)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82 establishes statutory liability and recourse; it does not address negotiated contractual caps.",
   "segment": null
  },
  {
   "row_id": "citation:154",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, Article 83",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/83",
   "notes": "Article 83 provides for regulatory fines; a negotiated contractual cap is distinct from those fines.",
   "segment": null
  },
  {
   "row_id": "citation:155",
   "case_id": "CG2",
   "kind": "citation",
   "title": "6 Del. C. §12A-107(a), (d)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 12A-107(a) recognizes electronic signatures; subsection (d) permits them to satisfy legal signature requirements.",
   "segment": null
  },
  {
   "row_id": "citation:156",
   "case_id": "CG1",
   "kind": "citation",
   "title": "YC post-money SAFE, §5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The standard post-money SAFE specifies conditions for majority amendments; the executed form controls.",
   "segment": null
  },
  {
   "row_id": "citation:157",
   "case_id": "CG3",
   "kind": "citation",
   "title": "N.Y. General Obligations Law § 5-1103",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The writing must be signed by the party against whom enforcement is sought or that party’s agent.",
   "segment": null
  },
  {
   "row_id": "citation:158",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. § 1839(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "The definition also requires that the information not be readily ascertainable through proper means by qualifying persons.",
   "segment": null
  },
  {
   "row_id": "citation:159",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Section 141(f) of the Delaware General Corporation Law",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Subject to restrictions in the certificate of incorporation or bylaws; electronic consents may be filed with corporate records.",
   "segment": null
  },
  {
   "row_id": "citation:160",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) permits unanimous written or electronic consent, subject to certificate or bylaw restrictions.",
   "segment": null
  },
  {
   "row_id": "citation:161",
   "case_id": "CG4",
   "kind": "citation",
   "title": "8 Del. C. §116(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 116(a) permits electronic documentation and signatures, subject to exceptions in subsection (b).",
   "segment": null
  },
  {
   "row_id": "citation:162",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Y Combinator, U.S. post-money SAFE, \"Valuation Cap, no Discount,\" §§1 to 2",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Sections 1 to 2 cover conversion events and relevant capitalization definitions.",
   "segment": null
  },
  {
   "row_id": "citation:163",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Securities Act of 1933 §4(a)(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6-astra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/15/77d",
   "notes": "Section 4(a)(2) exempts issuer transactions not involving a public offering; application depends on the facts.",
   "segment": null
  },
  {
   "row_id": "citation:164",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 33(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Notification is required unless risk is unlikely; the 72-hour deadline applies where feasible, with reasons required for delay.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:165",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 33(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) states the proposition verbatim; the repeated (2) appears redundant.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:166",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(2) supports the proposition; the repeated '(2)' is redundant.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:167",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(a) requires documented controller instructions, subject to an exception for legally required processing.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:168",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(e)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Assistance is required insofar as possible, taking into account the nature of processing.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:169",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Assistance accounts for the nature of processing and information available to the processor; it must be contractually stipulated.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:170",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(g)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The controller chooses deletion or return; existing copies must also be deleted unless applicable law requires storage.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:171",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(h)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(h) requires these processor obligations to be stipulated in the contract or other legal act.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:172",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(a)-(h)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(a) to (h) specifies mandatory processor terms; the introductory text also requires other contract particulars.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:173",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Chapter V",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Chapter V governs transfers to third countries or international organisations, including adequacy, safeguards and derogations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:174",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 83",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/83",
   "notes": "Article 83 sets fine conditions and maximums; the source does not establish procurement teams’ negotiating practices.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:175",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Articles 32 to 36",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Articles 32 to 36 also address prior consultation; specific obligations differ between controllers and processors.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:176",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Article 45",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/45",
   "notes": "EU GDPR Article 45 permits adequacy-based transfers; the supplied page shows the UK version, where Article 45 was omitted.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:177",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision adopts the 2021 standard contractual clauses for personal data transfers to third countries.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:178",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2023/1795",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Decision 2023/1795 establishes adequacy for the EU-U.S. Data Privacy Framework under GDPR Article 45.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:179",
   "case_id": "03",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA) Art. 30",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30 mandates incident assistance; enhanced audit rights apply to ICT services supporting critical or important functions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:180",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The initial notification is generally due within four hours of classification as major and no later than 24 hours after awareness.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:181",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR Part 255",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The guides address unexpected material connections and potential liability for deceptive or unsubstantiated endorsements, not all creator statements.",
   "segment": null
  },
  {
   "row_id": "citation:182",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR s.255.1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.1",
   "notes": "Section 255.1 addresses honesty and liability for undisclosed unexpected material connections; disclosure is more specifically addressed in § 255.5.",
   "segment": null
  },
  {
   "row_id": "citation:183",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR s.255.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.5",
   "notes": "Section 255.5 addresses disclosure of unexpected material connections; advertiser liability is addressed elsewhere in the Guides.",
   "segment": null
  },
  {
   "row_id": "citation:184",
   "case_id": "06",
   "kind": "citation",
   "title": "CAP Code rule 2.1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 2.1 requires obvious identification, but an upfront label is not invariably required if the advertising is otherwise obvious.",
   "segment": null
  },
  {
   "row_id": "citation:185",
   "case_id": "06",
   "kind": "citation",
   "title": "UK Code of Non-broadcast Advertising and Direct & Promotional Marketing (CAP Code, Edition 12), Section 2 (Rules 2.1 and 2.4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Advertising must be identifiable, but explicit labels are not universally required; Rule 2.4 addresses advertorials.",
   "segment": null
  },
  {
   "row_id": "citation:186",
   "case_id": "06",
   "kind": "citation",
   "title": "21 U.S.C. s.321(g)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/21/321",
   "notes": "It defines drugs as stated, but does not explicitly distinguish them from cosmetics; a product can be both.",
   "segment": null
  },
  {
   "row_id": "citation:187",
   "case_id": "06",
   "kind": "citation",
   "title": "21 U.S.C. s.321(i)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/21/321",
   "notes": "Section 321(i) defines cosmetics; drug definitions appear in (g). Products can be both, and approval is not addressed.",
   "segment": null
  },
  {
   "row_id": "citation:188",
   "case_id": "06",
   "kind": "citation",
   "title": "Digital Markets, Competition and Consumers Act 2024, Part 4, Chapter 1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Part 4, Chapter 1 prohibits unfair commercial practices, including certain practices that conceal commercial intent.",
   "segment": null
  },
  {
   "row_id": "citation:189",
   "case_id": "06",
   "kind": "citation",
   "title": "Late Payment of Commercial Debts (Interest) Act 1998, s.4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 4 sets the default 30-day period before statutory interest runs when no payment date is agreed.",
   "segment": null
  },
  {
   "row_id": "citation:190",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA), Art. 19",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_19.html",
   "notes": "Article 19 requires reporting within time limits established under Article 20.",
   "segment": null
  },
  {
   "row_id": "citation:191",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301, Art. 5(1)(a)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 5(1)(a) sets both deadlines for the initial notification.",
   "segment": null
  },
  {
   "row_id": "citation:192",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "PRA SS2/21",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SS2/21 covers outsourcing and third-party risk; the claimed immediate incident reporting requirement is unverified.",
   "segment": null
  },
  {
   "row_id": "citation:193",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "FCA SYSC 8",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SYSC 8 addresses outsourcing, but does not impose a general immediate-reporting requirement for all outsourcing matters.",
   "segment": null
  },
  {
   "row_id": "citation:194",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR), Art. 33(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Notification must be without undue delay, within 72 hours where feasible; low-risk breaches are exempt.",
   "segment": null
  },
  {
   "row_id": "citation:195",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR), Art. 33(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) states this requirement; the citation repeats the paragraph number.",
   "segment": null
  },
  {
   "row_id": "citation:196",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA), Art. 30(2)-(3)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(2)-(3) specifies mandatory terms but does not require uncapped supplier liability.",
   "segment": null
  },
  {
   "row_id": "citation:197",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR), Art. 28",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28 sets contract requirements but does not mandate uncapped supplier liability.",
   "segment": null
  },
  {
   "row_id": "citation:198",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR), Art. 82",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82 permits contribution according to responsibility, but each responsible party may owe the data subject the entire damage.",
   "segment": null
  },
  {
   "row_id": "citation:199",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA), Art. 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Assistance applies to ICT incidents related to the service provided to the financial entity.",
   "segment": null
  },
  {
   "row_id": "citation:200",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA), Art. 30(3)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(3) covers heightened exit and audit terms; subcontracting terms are in Article 30(2)(a).",
   "segment": null
  },
  {
   "row_id": "citation:201",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "EBA Guidelines on outsourcing arrangements (EBA/GL/2019/02)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The guidelines set contractual requirements for outsourcing arrangements, including those of EU credit institutions.",
   "segment": null
  },
  {
   "row_id": "citation:202",
   "case_id": "01",
   "kind": "citation",
   "title": "8 Del. C. s. 157",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 157(a) grants this authority, subject to the certificate of incorporation.",
   "segment": null
  },
  {
   "row_id": "citation:203",
   "case_id": "01",
   "kind": "citation",
   "title": "Y Combinator Post-Money SAFE (Valuation Cap, No Discount), Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) permits majority-in-interest amendments subject to specified conditions.",
   "segment": null
  },
  {
   "row_id": "citation:204",
   "case_id": "CG2",
   "kind": "citation",
   "title": "Del. Code Ann. tit. 6, s. 18-402",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=no",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 18-402 addresses LLC management, not the binding effect of signed amendments or waivers.",
   "segment": null
  },
  {
   "row_id": "citation:205",
   "case_id": "CG2",
   "kind": "citation",
   "title": "Y Combinator Post-Money SAFE Form, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Written consent is required; additional conditions apply to amendments or waivers approved by a majority-in-interest.",
   "segment": null
  },
  {
   "row_id": "citation:206",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code section 1950.5(h)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "The landlord must return the remaining deposit and provide an itemized statement within 21 calendar days.",
   "segment": null
  },
  {
   "row_id": "citation:207",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code section 1950.5(m)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Subdivision (m) supports both the statutory damages and burden-of-proof propositions.",
   "segment": null
  },
  {
   "row_id": "citation:208",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code section 1950.5(f)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Written notice is required within a reasonable time, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:209",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code section 1950.5(o)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Small claims jurisdiction is subject to applicable monetary limits.",
   "segment": null
  },
  {
   "row_id": "citation:210",
   "case_id": "18",
   "kind": "citation",
   "title": "Code of Civil Procedure section 116.221",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.221",
   "notes": "The $12,500 limit applies to natural persons, subject to statutory exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:211",
   "case_id": "18",
   "kind": "citation",
   "title": "Code of Civil Procedure section 116.530(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "The general prohibition applies, but statutory exceptions permit attorneys to appear in specified capacities.",
   "segment": null
  },
  {
   "row_id": "citation:212",
   "case_id": "18",
   "kind": "citation",
   "title": "Code of Civil Procedure section 116.530(c)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "Subdivision (c)(1) expressly permits attorney advice before or after commencement of the action.",
   "segment": null
  },
  {
   "row_id": "citation:213",
   "case_id": "18",
   "kind": "citation",
   "title": "Code of Civil Procedure sections 116.260 and 116.940",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.260",
   "notes": "Section 116.260 requires free assistance in each county as provided in section 116.940.",
   "segment": null
  },
  {
   "row_id": "citation:214",
   "case_id": "18",
   "kind": "citation",
   "title": "Code of Civil Procedure section 116.230(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.230",
   "notes": "Subdivision (b), not (a), sets fees by claim amount; subdivision (c) sets a fee for frequent filers.",
   "segment": null
  },
  {
   "row_id": "citation:215",
   "case_id": "CG1",
   "kind": "citation",
   "title": "6 Del. C. s. 2708",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 2708 generally permits Delaware governing-law clauses for transactions involving at least $100,000.",
   "segment": null
  },
  {
   "row_id": "citation:216",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. s. 1839(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "Section 1839(3)(A) requires reasonable measures to maintain secrecy; the repeated (3) is redundant.",
   "segment": null
  },
  {
   "row_id": "citation:217",
   "case_id": "CG3",
   "kind": "citation",
   "title": "Ashland Management Inc. v. Janien, 82 N.Y.2d 395, 407 (1993)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The Court applied the Restatement of Torts § 757, comment b, definition of a trade secret.",
   "segment": null
  },
  {
   "row_id": "citation:218",
   "case_id": "CG3",
   "kind": "citation",
   "title": "Defend Trade Secrets Act",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The DTSA provides federal trade-secret protection; the statutory definition appears at 18 U.S.C. § 1839(3).",
   "segment": null
  },
  {
   "row_id": "citation:219",
   "case_id": "CG3",
   "kind": "citation",
   "title": "Restatement of Torts s. 757",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The trade-secret definition appears in comment b to § 757 and was adopted by the New York Court of Appeals.",
   "segment": null
  },
  {
   "row_id": "citation:220",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL Section 141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) requires unanimous written or electronic consent for board action without a meeting.",
   "segment": null
  },
  {
   "row_id": "citation:221",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL Section 144",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 144(a) can cover a director’s SAFE investment, but disclosure is not universally required; alternative protections exist.",
   "segment": null
  },
  {
   "row_id": "citation:222",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Securities Act of 1933",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The Act provides registration exemptions, but the stated reliance for these particular offerings cannot be verified.",
   "segment": null
  },
  {
   "row_id": "citation:223",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Rule 506(b) of Regulation D",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.506",
   "notes": "Rule 506(b) also permits sales to a limited number of sophisticated non-accredited investors, subject to conditions.",
   "segment": null
  },
  {
   "row_id": "citation:224",
   "case_id": "CG4",
   "kind": "citation",
   "title": "17 CFR 230.503(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.503",
   "notes": "Applies to Rule 504 or 506 offerings; weekends and holidays can extend the deadline. See (a)(1).",
   "segment": null
  },
  {
   "row_id": "citation:225",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Rule 501(a) of Regulation D",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.501",
   "notes": "Rule 501(a) defines accredited investor; it does not establish whether particular SAFE purchasers qualify.",
   "segment": null
  },
  {
   "row_id": "citation:226",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 6(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/6/",
   "notes": "Article 6(2) covers Annex III use cases, but Article 6(1) also classifies certain Annex I product-related systems as high-risk.",
   "segment": null
  },
  {
   "row_id": "citation:227",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Annex III",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III lists these uses, but other high-risk systems fall under Annex I; exceptions also apply.",
   "segment": null
  },
  {
   "row_id": "citation:228",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Annex III, point 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III, point 4, covers specified employment, recruitment, and candidate-evaluation AI systems.",
   "segment": null
  },
  {
   "row_id": "citation:229",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Annex III, point 5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III, point 5(b), covers creditworthiness assessments, except AI systems used to detect financial fraud.",
   "segment": null
  },
  {
   "row_id": "citation:230",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Annex III point 1(c)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III, point 1(c), lists emotion recognition systems as a high-risk use.",
   "segment": null
  },
  {
   "row_id": "citation:231",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 50(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Applies to systems intended to interact directly; a qualified law-enforcement exception also applies.",
   "segment": null
  },
  {
   "row_id": "citation:232",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 50(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(2) imposes this obligation on providers, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:233",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/4/",
   "notes": "Article 4 requires measures to ensure sufficient literacy to the best extent, not a guaranteed level for every individual.",
   "segment": null
  },
  {
   "row_id": "citation:234",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/113/",
   "notes": "Article 113 sets 2 August 2026 as the general application date, subject to exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:235",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Article 5 prohibits specified manipulative or deceptive practices and social scoring, subject to statutory conditions.",
   "segment": null
  },
  {
   "row_id": "citation:236",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 5(1)(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Exceptions apply where the system is intended for medical or safety reasons.",
   "segment": null
  },
  {
   "row_id": "citation:237",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Art. 99(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/99/",
   "notes": "Article 99(4)(g) covers Article 50 transparency obligations; the higher threshold applies to undertakings, subject to the SME exception.",
   "segment": null
  },
  {
   "row_id": "citation:238",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689 (EU AI Act), Title VIII",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=no; accurate=no",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The Act uses chapters, not titles. General-purpose AI model provider obligations are in Chapter V.",
   "segment": null
  },
  {
   "row_id": "citation:239",
   "case_id": "13",
   "kind": "citation",
   "title": "Directive 2002/58/EC (ePrivacy Directive), Art. 13",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eudr/2002/58/article/13",
   "notes": "Prior consent is required for specified methods, but an existing-customer exception and opt-out options apply in some cases.",
   "segment": null
  },
  {
   "row_id": "citation:240",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR), Art. 14",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/14",
   "notes": "Article 14 requires controllers to provide information when data are obtained indirectly, subject to exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:241",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR), Art. 21",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "google/gemini-4-argon",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/21",
   "notes": "Article 21(2) expressly provides this right.",
   "segment": null
  },
  {
   "row_id": "citation:242",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Art. 28",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28 sets processor requirements, including mandatory contractual terms; the DPA’s compliance cannot be assessed without its terms.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:243",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(2) expressly permits general written authorisation subject to notice and an opportunity to object.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:244",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(3)(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The contract must stipulate this requirement, subject to an exception where domestic law requires processing.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:245",
   "case_id": "03",
   "kind": "citation",
   "title": "final subparagraph of Art. 28(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The final subparagraph requires immediate notification when the processor considers an instruction infringes applicable data protection law.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:246",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(3)(g)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The contract must also require deletion of existing copies unless domestic law requires storage.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:247",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(3)(h)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(h) requires the processing contract or legal act to stipulate this audit obligation.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:248",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(4) requires flow-down and makes the initial processor fully liable to the controller for the sub-processor’s obligations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:249",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 33(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(1) applies unless risk is unlikely; 72 hours runs from awareness and applies where feasible.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:250",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 33(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) expressly imposes this duty for personal data breaches; the second (2) is redundant.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:251",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 33(3) of Regulation (EU) 2016/679",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(3) specifies the minimum content of a breach notification to the supervisory authority.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:252",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 33(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(4) permits phased provision where simultaneous provision is impossible, without undue further delay. The citation repeats (4).",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:253",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 82 of Regulation (EU) 2016/679",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82 provides compensation for material or non-material damage and sets out controller and processor liability.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:254",
   "case_id": "03",
   "kind": "citation",
   "title": "Articles 32 to 36 of Regulation (EU) 2016/679",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Articles 32 to 36 cover security, breach notification, impact assessments, and prior consultation.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:255",
   "case_id": "03",
   "kind": "citation",
   "title": "Chapter V of Regulation (EU) 2016/679",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Chapter V requires a lawful basis for transfers to third countries, such as adequacy, safeguards, or an applicable derogation.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:256",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision adopts standard contractual clauses for transfers of personal data to third countries under the GDPR.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:257",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2023/1795",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision grants adequacy for transfers to participating US organizations certified under the EU-US Data Privacy Framework.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:258",
   "case_id": "03",
   "kind": "citation",
   "title": "Clause 14 of the Standard Contractual Clauses",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Clause 14 requires an assessment of relevant local laws and practices affecting compliance.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:259",
   "case_id": "03",
   "kind": "citation",
   "title": "Clause 9(a) of the Standard Contractual Clauses",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "In the 2021 EU SCCs, Clause 9(a) sets authorization options and notice periods for sub-processor changes.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:260",
   "case_id": "03",
   "kind": "citation",
   "title": "Clause 17 of the Standard Contractual Clauses",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Clause 17 of the 2021 EU Standard Contractual Clauses sets out governing law options.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:261",
   "case_id": "03",
   "kind": "citation",
   "title": "DORA Art. 30",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Incident assistance terms are generally mandatory; audit rights are specifically required for ICT services supporting critical or important functions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:262",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR 255.1(d)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.1",
   "notes": "Advertisers are subject to liability for misleading endorsements and undisclosed unexpected material connections, not all undisclosed connections.",
   "segment": null
  },
  {
   "row_id": "citation:263",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR 255.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.5",
   "notes": "Disclosure is required when a material connection is not reasonably expected by the audience.",
   "segment": null
  },
  {
   "row_id": "citation:264",
   "case_id": "06",
   "kind": "citation",
   "title": "15 U.S.C. s.45",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/15/45",
   "notes": "Section 45(a)(1) declares unfair or deceptive acts or practices in or affecting commerce unlawful.",
   "segment": null
  },
  {
   "row_id": "citation:265",
   "case_id": "06",
   "kind": "citation",
   "title": "21 U.S.C. s.321(g)(1), (i)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/21/321",
   "notes": "The definitions support drug classification based on intended treatment use; the statute does not itself require escalation.",
   "segment": null
  },
  {
   "row_id": "citation:266",
   "case_id": "06",
   "kind": "citation",
   "title": "CAP Code section 2",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "CAP Code rule 2.1 requires marketing communications to be obviously identifiable as such.",
   "segment": null
  },
  {
   "row_id": "citation:267",
   "case_id": "06",
   "kind": "citation",
   "title": "CAP Code section 12",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 12 covers advertising claims for beauty products, among other health-related products.",
   "segment": null
  },
  {
   "row_id": "citation:268",
   "case_id": "06",
   "kind": "citation",
   "title": "Digital Markets, Competition and Consumers Act 2024, Part 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Part 4 addresses unfair commercial practices, including undisclosed paid promotional content.",
   "segment": null
  },
  {
   "row_id": "citation:269",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "DORA (Regulation (EU) 2022/2554)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "DORA applies to covered financial entities, including credit institutions, from 17 January 2025.",
   "segment": null
  },
  {
   "row_id": "citation:270",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 28 (DORA)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 28 requires financial entities, including banks, to manage ICT third-party risk, including risks involving ICT service providers.",
   "segment": null
  },
  {
   "row_id": "citation:271",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 30 (DORA)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30 specifies mandatory contractual content but does not establish a liability standard.",
   "segment": null
  },
  {
   "row_id": "citation:272",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 30(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=no; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(2) sets minimum terms, but the additional ‘(2)’ does not identify an existing subsection.",
   "segment": null
  },
  {
   "row_id": "citation:273",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Requires contractual assistance for ICT incidents related to the provided service, free or at a predetermined cost.",
   "segment": null
  },
  {
   "row_id": "citation:274",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 30(3)(b)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(3)(b) requires contractual notice periods and provider reporting obligations for ICT services supporting critical or important functions.",
   "segment": null
  },
  {
   "row_id": "citation:275",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 19 (DORA)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_19.html",
   "notes": "Financial entities remain responsible even if they outsource reporting to a third-party provider under Article 19(5).",
   "segment": null
  },
  {
   "row_id": "citation:276",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Art. 2(1) (DORA)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_2.html",
   "notes": "Article 2(1) lists entities within DORA’s scope, subject to exceptions in paragraphs 3 and 4.",
   "segment": null
  },
  {
   "row_id": "citation:277",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Delegated Regulation 2025/301 Art. 5",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The initial notification is due within four hours of classification and no later than 24 hours after awareness.",
   "segment": null
  },
  {
   "row_id": "citation:278",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301 Art. 5(1)(a) to (c)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The 72 hours run from initial notification; the month runs from the latest intermediate report.",
   "segment": null
  },
  {
   "row_id": "citation:279",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 33(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The controller must generally notify the supervisory authority within 72 hours of becoming aware of the breach.",
   "segment": null
  },
  {
   "row_id": "citation:280",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 33(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 33(2) requires notification without undue delay after the processor becomes aware of the breach.",
   "segment": null
  },
  {
   "row_id": "citation:281",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Article 33(3) GDPR",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 33(3) lists required information; Article 33(4) governs providing information in phases.",
   "segment": null
  },
  {
   "row_id": "citation:282",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Article 34 GDPR",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 34 governs communication of personal data breaches to affected data subjects.",
   "segment": null
  },
  {
   "row_id": "citation:283",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 28(3)(f) requires assistance with Articles 32 to 36, including security and breach obligations.",
   "segment": null
  },
  {
   "row_id": "citation:284",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 82(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Processors may also be liable for acting outside or contrary to the controller’s lawful instructions.",
   "segment": null
  },
  {
   "row_id": "citation:285",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "PRA SS2/21",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SS2/21 addresses outsourcing and third-party risk; the claimed absence of a supplier deadline is unverified.",
   "segment": null
  },
  {
   "row_id": "citation:286",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "FCA Handbook SYSC 8",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SYSC 8 requires disclosure of material developments, but does not expressly require incident-notification terms or specify a supplier deadline.",
   "segment": null
  },
  {
   "row_id": "citation:287",
   "case_id": "CG2",
   "kind": "citation",
   "title": "6 Del. C. § 12A-107",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 12A-107 recognizes the legal effect of electronic records and signatures.",
   "segment": null
  },
  {
   "row_id": "citation:288",
   "case_id": "CG2",
   "kind": "citation",
   "title": "15 U.S.C. § 7001(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=no",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/15/7001",
   "notes": "Section 7001(a) prohibits denying legal effect solely because signatures or records are electronic; the proposition reverses this rule.",
   "segment": null
  },
  {
   "row_id": "citation:289",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code s.1950.5(h)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Within 21 days, the landlord must provide an itemized statement and return any remaining deposit.",
   "segment": null
  },
  {
   "row_id": "citation:290",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code s.1950.5(m)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Subdivision (m) permits statutory damages up to twice the security deposit for bad-faith retention.",
   "segment": null
  },
  {
   "row_id": "citation:291",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Civ. Code s.1950.5(o)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=unsure; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "The provided source text ends before subdivision (o), so its existence and contents cannot be verified.",
   "segment": null
  },
  {
   "row_id": "citation:292",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.221",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.221",
   "notes": "Section 116.221 sets the $12,500 limit for natural persons, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:293",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.530",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "The general prohibition has exceptions, including attorneys appearing for themselves and representing parties on appeal.",
   "segment": null
  },
  {
   "row_id": "citation:294",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.940",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.940",
   "notes": "Section 116.940 specifies advisory services, but does not itself state they are free; Section 116.260 requires the service.",
   "segment": null
  },
  {
   "row_id": "citation:295",
   "case_id": "18",
   "kind": "citation",
   "title": "Granberry v. Islay Investments (1995) 9 Cal.4th 738",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The court permitted proven damage deductions despite the landlord’s failure to provide a timely itemized statement.",
   "segment": null
  },
  {
   "row_id": "citation:296",
   "case_id": "18",
   "kind": "citation",
   "title": "Form SC-100 (Plaintiff's Claim)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SC-100 is California’s official Plaintiff’s Claim and ORDER to Go to Small Claims Court form.",
   "segment": null
  },
  {
   "row_id": "citation:297",
   "case_id": "18",
   "kind": "citation",
   "title": "Form FW-001 (fee waiver)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Form FW-001 is California’s official Request to Waive Court Fees.",
   "segment": null
  },
  {
   "row_id": "citation:298",
   "case_id": "18",
   "kind": "citation",
   "title": "Form SC-104 (Proof of Service)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SC-104 is California’s Proof of Service form for small claims cases.",
   "segment": null
  },
  {
   "row_id": "citation:299",
   "case_id": "CG1",
   "kind": "citation",
   "title": "YC Post-Money SAFE, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) also requires the Company’s written consent; investor or majority consent alone is insufficient.",
   "segment": null
  },
  {
   "row_id": "citation:300",
   "case_id": "CG1",
   "kind": "citation",
   "title": "6 Del. C. § 2708",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Generally permits Delaware choice of law for transactions involving at least $100,000, subject to exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:301",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. §1839(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "Section 1839(3) defines trade secret; the agreement’s use of that definition cannot be verified from the source.",
   "segment": null
  },
  {
   "row_id": "citation:302",
   "case_id": "CG3",
   "kind": "citation",
   "title": "15 U.S.C. §7001 (ESIGN)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/15/7001",
   "notes": "Section 7001(a) states this rule for transactions affecting interstate or foreign commerce, subject to statutory exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:303",
   "case_id": "CG3",
   "kind": "citation",
   "title": "N.Y. State Technology Law §§301 to 309 (ESRA)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "ESRA governs the validity of electronic signatures; section 304 addresses their legal effect.",
   "segment": null
  },
  {
   "row_id": "citation:304",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL §141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) requires unanimous written or electronic consent for board action without a meeting.",
   "segment": null
  },
  {
   "row_id": "citation:305",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Section 141(f) of the Delaware General Corporation Law",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) permits unanimous written consent unless restricted by the certificate of incorporation or bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:306",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL §144",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 144 governs interested transactions; its protections require additional conditions, including disclosure and approval without gross negligence.",
   "segment": null
  },
  {
   "row_id": "citation:307",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL §157",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 157 authorizes stock rights, but the text does not establish whether SAFE classification is unsettled.",
   "segment": null
  },
  {
   "row_id": "citation:308",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Rule 501(a) of Regulation D under the Securities Act of 1933",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.501",
   "notes": "Rule 501(a) defines accredited investor; it does not itself limit SAFE purchases or establish the offering’s terms.",
   "segment": null
  },
  {
   "row_id": "citation:309",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Rule 506(b) of Regulation D",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.506",
   "notes": "Rule 506(b) permits offerings to accredited investors without general solicitation; certain non-accredited investors may also participate.",
   "segment": null
  },
  {
   "row_id": "citation:310",
   "case_id": "CG4",
   "kind": "citation",
   "title": "17 CFR 230.506(b)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.506",
   "notes": "Rule 506(b) prohibits general solicitation but permits up to 35 qualifying non-accredited purchasers; accreditation is not universally required.",
   "segment": null
  },
  {
   "row_id": "citation:311",
   "case_id": "CG4",
   "kind": "citation",
   "title": "17 CFR 230.503(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.503",
   "notes": "Applies to Rule 504 or 506 offerings; weekends and holidays may extend the deadline. The pinpoint is (a)(1).",
   "segment": null
  },
  {
   "row_id": "citation:312",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Form D",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Form D is generally due within 15 days after the first sale when relying on a Regulation D exemption.",
   "segment": null
  },
  {
   "row_id": "citation:313",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Securities Act of 1933",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SAFEs generally are securities subject to registration unless an exemption applies.",
   "segment": null
  },
  {
   "row_id": "citation:314",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Regulation (EU) 2024/1689 is the EU AI Act and covers the stated subjects.",
   "segment": null
  },
  {
   "row_id": "citation:315",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 50(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(1) imposes the obligation on providers for direct interaction, with an additional law-enforcement exception.",
   "segment": null
  },
  {
   "row_id": "citation:316",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 50(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(2) supports the rule, but provides exceptions; technical feasibility qualifies the effectiveness and reliability of solutions.",
   "segment": null
  },
  {
   "row_id": "citation:317",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 50(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(4) establishes the disclosure obligation, subject to law-enforcement and artistic-content qualifications.",
   "segment": null
  },
  {
   "row_id": "citation:318",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 50(5)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(5) applies to information under paragraphs 1, 3 and 4, not all Article 50 obligations.",
   "segment": null
  },
  {
   "row_id": "citation:319",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 3(60)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/3/",
   "notes": "The definition also requires that the content would falsely appear authentic or truthful to a person.",
   "segment": null
  },
  {
   "row_id": "citation:320",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/4/",
   "notes": "Article 4 requires measures to ensure sufficient AI literacy to the best extent, considering relevant circumstances.",
   "segment": null
  },
  {
   "row_id": "citation:321",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Article 5(1)(a) to (b) prohibits these practices subject to specified behavioral distortion and harm conditions.",
   "segment": null
  },
  {
   "row_id": "citation:322",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 5(1)(a)-(b)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Manipulative or deceptive techniques must be purposeful; vulnerabilities must arise from specified characteristics or circumstances.",
   "segment": null
  },
  {
   "row_id": "citation:323",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 6(1)-(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/6/",
   "notes": "Product cases also require third-party conformity assessment; Annex III classification is subject to exceptions under Article 6(3).",
   "segment": null
  },
  {
   "row_id": "citation:324",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/113/",
   "notes": "Article 113 specifies both dates, including 2 February 2025 for Chapters I and II.",
   "segment": null
  },
  {
   "row_id": "citation:325",
   "case_id": "13",
   "kind": "citation",
   "title": "Art. 99(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/99/",
   "notes": "Article 99(4)(g) covers Article 50; the higher cap applies to undertakings, subject to SME exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:326",
   "case_id": "13",
   "kind": "citation",
   "title": "Annex I",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Assuming the EU AI Act: high-risk classification also requires third-party conformity assessment under Article 6(1).",
   "segment": null
  },
  {
   "row_id": "citation:327",
   "case_id": "13",
   "kind": "citation",
   "title": "Annex III",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Assuming the EU AI Act, Annex III lists these high-risk AI use areas.",
   "segment": null
  },
  {
   "row_id": "citation:328",
   "case_id": "13",
   "kind": "citation",
   "title": "Annex III 1(c)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "EU AI Act Annex III, point 1(c), lists emotion recognition systems under biometrics, insofar as their use is permitted.",
   "segment": null
  },
  {
   "row_id": "citation:329",
   "case_id": "13",
   "kind": "citation",
   "title": "Annex III 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "EU AI Act Annex III, point 4 covers specified employment uses; not every employment-related AI system is high-risk.",
   "segment": null
  },
  {
   "row_id": "citation:330",
   "case_id": "13",
   "kind": "citation",
   "title": "GDPR Art. 21(2)-(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/21",
   "notes": "Article 21(2) to (3) expressly provides this right and requires direct marketing processing to cease upon objection.",
   "segment": null
  },
  {
   "row_id": "citation:331",
   "case_id": "13",
   "kind": "citation",
   "title": "Directive 2002/58/EC Art. 13",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "pascal-1-fallback",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eudr/2002/58/article/13",
   "notes": "Prior consent is generally required for email marketing; a limited existing-customer exception permits an opt-out.",
   "segment": null
  },
  {
   "row_id": "citation:332",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The GDPR addresses all listed topics across multiple provisions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:333",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28(4) of the GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(4) also permits another legal act, not only a contract; the initial processor remains liable to the controller.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:334",
   "case_id": "03",
   "kind": "citation",
   "title": "Article 28 of the GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(h) requires these obligations to be stipulated in the processor’s contract or other legal act.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:335",
   "case_id": "03",
   "kind": "citation",
   "title": "Chapter V of the GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Chapter V governs transfers to third countries and international organisations, including adequacy decisions, safeguards, and derogations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:336",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision adopts SCCs, but their use is not mandatory for all international transfers; other safeguards may be available.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:337",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Articles 28(2) to (4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Articles 28(2) to (4) cover instructions, audits, sub-processor authorization, and obligations.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:338",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Articles 33(1) to (2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(1) addresses controller reporting; 33(2) requires processor notice to the controller without undue delay.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:339",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Articles 44 to 46",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Articles 44 to 46 govern general transfer principles, adequacy decisions, and appropriate safeguards.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:340",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914, Annex, Modules Two and Three",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Modules Two and Three cover controller-to-processor and processor-to-processor transfers, respectively.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:341",
   "case_id": "03",
   "kind": "citation",
   "title": "GDPR Article 82",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82 provides compensation rights, but does not expressly address whether a contractual liability cap can limit them.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:342",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR § 255.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.5",
   "notes": "Disclosure is required when the material connection is not reasonably expected by the audience.",
   "segment": null
  },
  {
   "row_id": "citation:343",
   "case_id": "06",
   "kind": "citation",
   "title": "UK CAP Code, rule 2.1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 2.1 requires marketing communications to be obviously identifiable as such.",
   "segment": null
  },
  {
   "row_id": "citation:344",
   "case_id": "06",
   "kind": "citation",
   "title": "UK CAP Code, rule 2.4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=no",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 2.4 concerns identifying advertorials as marketing communications, not disclosing the marketer’s identity.",
   "segment": null
  },
  {
   "row_id": "citation:345",
   "case_id": "06",
   "kind": "citation",
   "title": "UK CAP Code, rule 3.7",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Rule 3.7 requires documentary evidence for objectively substantiable claims before distribution or submission for publication.",
   "segment": null
  },
  {
   "row_id": "citation:346",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, art. 28",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 28 includes termination requirements; Article 30 specifies additional mandatory contractual terms.",
   "segment": null
  },
  {
   "row_id": "citation:347",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, art. 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "The incident must relate to the ICT service provided to the financial entity.",
   "segment": null
  },
  {
   "row_id": "citation:348",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554, art. 30(3)(b)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(3)(b) requires notice and reporting obligations; incident assistance is separately required under Article 30(2)(f).",
   "segment": null
  },
  {
   "row_id": "citation:349",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301, art. 5(1)(a)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 5(1)(a) sets the deadline for initial notification of a major ICT-related incident.",
   "segment": null
  },
  {
   "row_id": "citation:350",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, arts. 33(1) to (2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Controller notification is not required where the breach is unlikely to risk individuals’ rights and freedoms.",
   "segment": null
  },
  {
   "row_id": "citation:351",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR, art. 82",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82(1) expressly provides this right where damage results from an infringement.",
   "segment": null
  },
  {
   "row_id": "citation:352",
   "case_id": "01",
   "kind": "citation",
   "title": "YC post-money SAFE, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 5(a) addresses amendments; any specific MFN election and restatement terms must also be considered.",
   "segment": null
  },
  {
   "row_id": "citation:353",
   "case_id": "18",
   "kind": "citation",
   "title": "California Civil Code § 1950.5(h)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Section 1950.5 exists, but the supplied text does not establish whether subsection (h)(1) contains this rule.",
   "segment": null
  },
  {
   "row_id": "citation:354",
   "case_id": "18",
   "kind": "citation",
   "title": "California Civil Code § 1950.5(m)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "Subdivision (m) addresses the landlord’s burden; bad-faith retention damages are addressed in subdivision (l).",
   "segment": null
  },
  {
   "row_id": "citation:355",
   "case_id": "18",
   "kind": "citation",
   "title": "California Code of Civil Procedure § 116.221",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.221",
   "notes": "Section 116.221 sets a $12,500 limit for actions brought by natural persons, subject to statutory exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:356",
   "case_id": "18",
   "kind": "citation",
   "title": "California Code of Civil Procedure § 116.230",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.230",
   "notes": "Section 116.230 specifies small claims filing fees based on claim amount and prior filings.",
   "segment": null
  },
  {
   "row_id": "citation:357",
   "case_id": "18",
   "kind": "citation",
   "title": "California Code of Civil Procedure § 116.340",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.340",
   "notes": "Section 116.340 governs service of the small claims claim and order on the defendant.",
   "segment": null
  },
  {
   "row_id": "citation:358",
   "case_id": "18",
   "kind": "citation",
   "title": "California Code of Civil Procedure § 116.530",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "Section generally prohibits attorney participation in conducting or defending small claims actions, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:359",
   "case_id": "18",
   "kind": "citation",
   "title": "California Code of Civil Procedure § 116.940",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.940",
   "notes": "Section 116.940 specifies small claims advisory services and adviser qualifications, subject to a limited county exemption.",
   "segment": null
  },
  {
   "row_id": "citation:360",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. § 1839(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "Section 1839(3) defines trade secrets but does not expressly specify a protection period; the citation repeats (3).",
   "segment": null
  },
  {
   "row_id": "citation:361",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Section 141(f) of the Delaware General Corporation Law",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) permits unanimous written consent unless restricted by the certificate of incorporation or bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:362",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Delaware General Corporation Law § 141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 141(f) permits unanimous written consent unless restricted by the certificate of incorporation or bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:363",
   "case_id": "CG4",
   "kind": "citation",
   "title": "Delaware General Corporation Law § 116",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 116 addresses electronic documents, signatures, and delivery, subject to specified exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:364",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 2",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/2/",
   "notes": "Article 2 sets out scope, including specified providers and deployers located outside the EU.",
   "segment": null
  },
  {
   "row_id": "citation:365",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 3",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/3/",
   "notes": "Article 3 defines provider, deployer, importer, distributor and other roles relevant to allocating obligations.",
   "segment": null
  },
  {
   "row_id": "citation:366",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/6/",
   "notes": "Article 6 establishes high-risk classification rules, including criteria concerning intended use and product safety.",
   "segment": null
  },
  {
   "row_id": "citation:367",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Annex III",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Annex III lists high-risk use cases, including recruitment screening and creditworthiness assessment.",
   "segment": null
  },
  {
   "row_id": "citation:368",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/4/",
   "notes": "Article 4 requires measures to ensure sufficient AI literacy to providers’ and deployers’ best extent, not a guaranteed outcome.",
   "segment": null
  },
  {
   "row_id": "citation:369",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Article 5(1)(a) and (b) prohibit specified manipulative and exploitative AI practices subject to additional conditions.",
   "segment": null
  },
  {
   "row_id": "citation:370",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 50",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(1), (2), and (4) establish these obligations, subject to qualifications and exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:371",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/113/",
   "notes": "Article 113 sets the general application date and earlier dates for specified chapters.",
   "segment": null
  },
  {
   "row_id": "citation:372",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Art. 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/6",
   "notes": "Article 6 sets out lawful bases; the supplied source is the amended UK version.",
   "segment": null
  },
  {
   "row_id": "citation:373",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Art. 14",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/14",
   "notes": "Article 14 requires such information, subject to exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:374",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Art. 21",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/21",
   "notes": "Article 21(2) expressly provides this right.",
   "segment": null
  },
  {
   "row_id": "citation:375",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Art. 28",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28 requires binding processor terms; the contract or other legal act must be in writing, including electronic form.",
   "segment": null
  },
  {
   "row_id": "citation:376",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Art. 44",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/44",
   "notes": "Article 44 sets the general principle, but transfers may also rely on adequacy decisions or applicable derogations.",
   "segment": null
  },
  {
   "row_id": "citation:377",
   "case_id": "13",
   "kind": "citation",
   "title": "Directive 2002/58/EC, Art. 13",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "openai/gpt-6.1-terra",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eudr/2002/58/article/13",
   "notes": "Article 13 regulates unsolicited direct marketing and requires Member States to adopt specified protections and remedies.",
   "segment": null
  },
  {
   "row_id": "citation:378",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(3)(a) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Requires a contractual or legal stipulation; an exception applies where processing is required by law.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:379",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(3) (final subparagraph) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "The processor must immediately inform the controller if, in its opinion, an instruction infringes applicable data protection law.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:380",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(2) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(2) expressly permits prior general written authorisation with notice of intended additions or replacements and opportunity to object.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:381",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(4) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(4) requires the same data protection obligations to be imposed on sub-processors by contract or other legal act.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:382",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 28(3)(h) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(h) requires these processor obligations to be stipulated in the contract or other legal act.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:383",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 33(1) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "The controller must notify without undue delay and, where feasible, within 72 hours, subject to the risk exception.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:384",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 33(2) GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) states this requirement expressly.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:385",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 82 GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82 grants compensation for qualifying damage, but does not expressly address contractual exclusion of liability.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:386",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 44 GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/44",
   "notes": "Article 44 states this principle in the EU GDPR; the supplied UK version notes its omission in 2026.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:387",
   "case_id": "03",
   "kind": "citation",
   "title": "Art. 46 GDPR",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/46",
   "notes": "Article 46 permits transfers using appropriate safeguards, including standard data protection clauses, subject to applicable conditions.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:388",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2021/914",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision adopts the 2021 Standard Contractual Clauses for transfers to third countries.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:389",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Implementing Decision (EU) 2023/1795",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The decision establishes adequacy for transfers to certified US organizations under the EU-US Data Privacy Framework.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:390",
   "case_id": "03",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA), Art. 30",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Article 30(2)(f) requires contractual assistance when a related ICT incident occurs.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:391",
   "case_id": "03",
   "kind": "citation",
   "title": "Commission Delegated Regulation (EU) 2025/301",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "It concerns DORA incident reporting, but relevance is not limited to cases where the customer is a financial entity.",
   "segment": "Regulated vendor data terms (customer-derived)"
  },
  {
   "row_id": "citation:392",
   "case_id": "06",
   "kind": "citation",
   "title": "15 U.S.C. s.45 (FTC Act s.5)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/15/45",
   "notes": "Section 5 prohibits deceptive practices; FTC substantiation doctrine requires reasonable evidence before making objective efficacy claims.",
   "segment": null
  },
  {
   "row_id": "citation:393",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR Part 255",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Part 255 includes FTC endorsement guides addressing disclosure of material connections.",
   "segment": null
  },
  {
   "row_id": "citation:394",
   "case_id": "06",
   "kind": "citation",
   "title": "16 CFR 255.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/16/255.5",
   "notes": "Disclosure applies when a material connection is not reasonably expected by the audience; expected payments may not require disclosure.",
   "segment": null
  },
  {
   "row_id": "citation:395",
   "case_id": "06",
   "kind": "citation",
   "title": "CAP Code, section 2",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 2 requires marketing communications to be identifiable; an explicit label is not always necessary.",
   "segment": null
  },
  {
   "row_id": "citation:396",
   "case_id": "06",
   "kind": "citation",
   "title": "Digital Markets, Competition and Consumers Act 2024, Part 4 Chapter 1",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=no",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Part 4 Chapter 1 addresses unfair commercial practices; the CMA’s direct consumer-law enforcement powers are in Part 3.",
   "segment": null
  },
  {
   "row_id": "citation:397",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2022/2554 (DORA)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "DORA includes ICT third-party contractual requirements and major incident reporting, and applies from 17 January 2025.",
   "segment": null
  },
  {
   "row_id": "citation:398",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "DORA Art. 28",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_28.html",
   "notes": "Article 28 sets general principles for ICT third-party risk management and related contractual arrangements.",
   "segment": null
  },
  {
   "row_id": "citation:399",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "DORA Art. 30(2)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.digital-operational-resilience-act.com/Article_30.html",
   "notes": "Assistance is required for ICT incidents related to the service provided to the financial entity.",
   "segment": null
  },
  {
   "row_id": "citation:400",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Delegated Regulation (EU) 2025/301, Art. 5",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Article 5 sets both deadlines for the initial notification of a major ICT-related incident.",
   "segment": null
  },
  {
   "row_id": "citation:401",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Delegated Regulation (EU) 2025/301, Art. 5(1)(a) to (c)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=unsure; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The regulation concerns ICT incident reporting, but I cannot verify the cited article’s deadlines.",
   "segment": null
  },
  {
   "row_id": "citation:402",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679 (GDPR)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The GDPR governs controller and processor duties, breach notification, and liability for personal data processing.",
   "segment": null
  },
  {
   "row_id": "citation:403",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 33(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Notification is required unless risk is unlikely; the 72-hour deadline applies where feasible, with reasons for delay otherwise.",
   "segment": null
  },
  {
   "row_id": "citation:404",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 33(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/33",
   "notes": "Article 33(2) requires notification without undue delay after the processor becomes aware of the breach.",
   "segment": null
  },
  {
   "row_id": "citation:405",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 28(3)(f)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/28",
   "notes": "Article 28(3)(f) covers assistance with Articles 32 to 36, including security and breach obligations.",
   "segment": null
  },
  {
   "row_id": "citation:406",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "GDPR Art. 82",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/82",
   "notes": "Article 82 provides compensation for material or non-material damage and specifies controller and processor liability.",
   "segment": null
  },
  {
   "row_id": "citation:407",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Unfair Contract Terms Act 1977",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "The Act regulates certain exclusion and limitation clauses, including through a reasonableness test in business contracts.",
   "segment": null
  },
  {
   "row_id": "citation:408",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Unfair Contract Terms Act 1977, s.2(1)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 2(1) prohibits excluding or restricting liability for death or personal injury resulting from negligence.",
   "segment": null
  },
  {
   "row_id": "citation:409",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Unfair Contract Terms Act 1977, s.2(2)",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 2(2) subjects exclusions or limitations of liability for other negligence loss or damage to reasonableness.",
   "segment": null
  },
  {
   "row_id": "citation:410",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "Unfair Contract Terms Act 1977, s.3",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Section 3 subjects specified liability restrictions in written standard terms to the reasonableness requirement.",
   "segment": null
  },
  {
   "row_id": "citation:411",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "PRA SS2/21, Outsourcing and third party risk management",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SS2/21 sets expectations for incident notification, audit and access rights, and exit arrangements in outsourcing agreements.",
   "segment": null
  },
  {
   "row_id": "citation:412",
   "case_id": "FG-L1",
   "kind": "citation",
   "title": "FCA Handbook SYSC 8",
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "SYSC 8 covers outsourcing oversight, access and termination, but does not impose a universal incident-notification requirement on all firms.",
   "segment": null
  },
  {
   "row_id": "citation:413",
   "case_id": "18",
   "kind": "citation",
   "title": "Civ. Code s.1950.5(h)(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=unsure; accurate=unsure",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "The supplied text confirms section 1950.5 exists but ends before subsection (h)(1).",
   "segment": null
  },
  {
   "row_id": "citation:414",
   "case_id": "18",
   "kind": "citation",
   "title": "Civil Code s.1950.5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_code_section_1950.5",
   "notes": "California Civil Code section 1950.5 authorizes statutory damages up to twice the security deposit for bad-faith retention.",
   "segment": null
  },
  {
   "row_id": "citation:415",
   "case_id": "18",
   "kind": "citation",
   "title": "CCP s.116.221",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.221",
   "notes": "The $12,500 limit applies to natural persons, subject to statutory exceptions.",
   "segment": null
  },
  {
   "row_id": "citation:416",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.530",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.530",
   "notes": "The general prohibition has exceptions, including attorneys appearing as parties or testifying as witnesses.",
   "segment": null
  },
  {
   "row_id": "citation:417",
   "case_id": "18",
   "kind": "citation",
   "title": "Cal. Code Civ. Proc. s.116.940",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://california.public.law/codes/ca_civ_proc_code_section_116.940",
   "notes": "Section governs advisory services, but does not specify they are free or at the court; some counties may exempt themselves from personal advice.",
   "segment": null
  },
  {
   "row_id": "citation:418",
   "case_id": "CG1",
   "kind": "citation",
   "title": "Y Combinator post-money SAFE, Section 5(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "These conditions apply to the majority-in-interest route; the company and individual investor may alternatively consent.",
   "segment": null
  },
  {
   "row_id": "citation:419",
   "case_id": "CG3",
   "kind": "citation",
   "title": "18 U.S.C. §1839(3)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/uscode/text/18/1839",
   "notes": "Section 1839(3) defines trade secret; the repeated (3) is a citation typo.",
   "segment": null
  },
  {
   "row_id": "citation:420",
   "case_id": "CG3",
   "kind": "citation",
   "title": "Restatement of Torts §757",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "New York courts have applied the trade-secret factors in § 757, comment b.",
   "segment": null
  },
  {
   "row_id": "citation:421",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL Section 141(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Subject to restrictions in the certificate of incorporation or bylaws.",
   "segment": null
  },
  {
   "row_id": "citation:422",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL Section 144",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Disinterested director approval is one path; informed disinterested stockholder approval or fairness can alternatively provide protection.",
   "segment": null
  },
  {
   "row_id": "citation:423",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL Sections 152 and 153",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Fully paid status requires receipt of consideration and compliance with Section 152; consideration may also be determined by stockholders or delegates.",
   "segment": null
  },
  {
   "row_id": "citation:424",
   "case_id": "CG4",
   "kind": "citation",
   "title": "DGCL Section 153",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://delcode.delaware.gov/title8/c001/",
   "notes": "Section 153 addresses consideration, but adequate consideration alone does not establish valid issuance or fully paid, non-assessable status.",
   "segment": null
  },
  {
   "row_id": "citation:425",
   "case_id": "CG4",
   "kind": "citation",
   "title": "17 CFR 230.503",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.law.cornell.edu/cfr/text/17/230.503",
   "notes": "Applies to offerings relying on Rules 504 or 506; weekend and holiday deadlines move to the next business day.",
   "segment": null
  },
  {
   "row_id": "citation:426",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2024/1689, Art. 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/6/",
   "notes": "Article 6 establishes classification rules for high-risk AI systems.",
   "segment": null
  },
  {
   "row_id": "citation:427",
   "case_id": "13",
   "kind": "citation",
   "title": "Annex III, point 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Assuming the EU AI Act, Annex III point 4(b) includes evaluating workers’ performance.",
   "segment": null
  },
  {
   "row_id": "citation:428",
   "case_id": "13",
   "kind": "citation",
   "title": "Annex III, point 5",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (model-checked)",
   "judge": "pascal-1",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Assuming the EU AI Act, point 5 covers credit scoring and insurance pricing specifically for life and health insurance.",
   "segment": null
  },
  {
   "row_id": "citation:429",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 50(1)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(1) supports the rule, but also provides a qualified exception for systems authorised for criminal law enforcement.",
   "segment": null
  },
  {
   "row_id": "citation:430",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 50(2)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(2) provides the marking obligation and an exception for standard editing assistance.",
   "segment": null
  },
  {
   "row_id": "citation:431",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 50(4)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/50/",
   "notes": "Article 50(4) covers deepfakes and certain public-interest text, not all generated or manipulated content; exceptions apply.",
   "segment": null
  },
  {
   "row_id": "citation:432",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 5(1)(a)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "Also requires purposeful manipulation or deception, material behavioural distortion and impaired informed decision-making; likely significant harm suffices.",
   "segment": null
  },
  {
   "row_id": "citation:433",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 5(1)(f)",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/5/",
   "notes": "The prohibition includes exceptions for medical or safety reasons.",
   "segment": null
  },
  {
   "row_id": "citation:434",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 4",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/4/",
   "notes": "Article 4 requires measures to ensure sufficient literacy to the providers’ and deployers’ best extent, not a guaranteed level.",
   "segment": null
  },
  {
   "row_id": "citation:435",
   "case_id": "13",
   "kind": "citation",
   "title": "Article 113",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://artificialintelligenceact.eu/article/113/",
   "notes": "Article 113 sets application dates, including general application from 2 August 2026.",
   "segment": null
  },
  {
   "row_id": "citation:436",
   "case_id": "13",
   "kind": "citation",
   "title": "Directive 2002/58/EC, Art. 13",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=partly",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eudr/2002/58/article/13",
   "notes": "Prior consent generally applies to automated calls and email; Member States may allow opt-out for other marketing calls.",
   "segment": null
  },
  {
   "row_id": "citation:437",
   "case_id": "13",
   "kind": "citation",
   "title": "Regulation (EU) 2016/679, Art. 6",
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": "anthropic/claude-haiku-5.5",
   "prompt_version": null,
   "run": "bakeoff citations",
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": "citation",
   "metric_value": "exists=yes; accurate=yes",
   "basis": "measured",
   "label": "MEASURED (fetched-checked)",
   "judge": "openai/gpt-6.1-terra",
   "run_date": "2026-10-31",
   "cost_usd": null,
   "latency_s": null,
   "link": "https://www.legislation.gov.uk/eur/2016/679/article/6",
   "notes": "Article 6 establishes lawful bases for processing; the supplied text is the amended UK version.",
   "segment": null
  },
  {
   "row_id": "task:02",
   "case_id": "02",
   "kind": "task",
   "title": "Terse term sheet + send intent",
   "source": "authored, fictional parties",
   "use_case": "Draft from terse ask",
   "jurisdiction": "DE;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/02-terse-term-sheet-and-safe.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:04",
   "case_id": "04",
   "kind": "task",
   "title": "Sponsorship review (customer-derived)",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": null,
   "notes": "Task file on disk; not yet run.",
   "segment": "Sports sponsorship (customer-derived)"
  },
  {
   "row_id": "task:07",
   "case_id": "07",
   "kind": "task",
   "title": "Delaware incorporation",
   "source": "authored, fictional parties",
   "use_case": "Formation",
   "jurisdiction": "DE;CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/07-delaware-incorporation.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:09",
   "case_id": "09",
   "kind": "task",
   "title": "Mutual NDA redline",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "CA;DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/09-mutual-nda-redline.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:10",
   "case_id": "10",
   "kind": "task",
   "title": "Contractor agreement + IP (California)",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/10-contractor-ip-california.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:12",
   "case_id": "12",
   "kind": "task",
   "title": "UK DSAR response",
   "source": "authored, fictional parties",
   "use_case": "Privacy request",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/12-uk-dsar-response.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:14",
   "case_id": "14",
   "kind": "task",
   "title": "Driver image rights (customer-derived)",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/14-driver-image-rights-england.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:15",
   "case_id": "15",
   "kind": "task",
   "title": "Dubai free zone vs mainland",
   "source": "authored, fictional parties",
   "use_case": "Formation",
   "jurisdiction": "AE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/15-dubai-free-zone-setup.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:16",
   "case_id": "16",
   "kind": "task",
   "title": "First employee in Singapore",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "SG",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/16-singapore-first-hire.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:17",
   "case_id": "17",
   "kind": "task",
   "title": "Indian subsidiary for engineers",
   "source": "authored, fictional parties",
   "use_case": "Formation",
   "jurisdiction": "IN;EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/17-india-subsidiary-engineers.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:19",
   "case_id": "19",
   "kind": "task",
   "title": "Unfair dismissal (England)",
   "source": "authored, fictional parties",
   "use_case": "Employment advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/19-unfair-dismissal-england.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:20",
   "case_id": "20",
   "kind": "task",
   "title": "Law firm NDA triage Associate",
   "source": "authored, fictional parties",
   "use_case": "Create Associate",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/20-law-firm-nda-triage-associate.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:21",
   "case_id": "21",
   "kind": "task",
   "title": "Renewal deadline tracker",
   "source": "authored, fictional parties",
   "use_case": "Obligations",
   "jurisdiction": "NY;DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/21-renewal-deadlines-tracker.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:24",
   "case_id": "24",
   "kind": "task",
   "title": "customer-derived venue hire (know limits)",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "MC",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/24-customer-derived-venue-hire-know-limits.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:25",
   "case_id": "25",
   "kind": "task",
   "title": "NY security deposit not returned",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/25-ny-security-deposit-tenant.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:26",
   "case_id": "26",
   "kind": "task",
   "title": "Late payment interest claim (England)",
   "source": "authored, fictional parties",
   "use_case": "Debt recovery maths",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/26-late-payment-interest-england.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:27",
   "case_id": "27",
   "kind": "task",
   "title": "Founder 83(b) election deadline",
   "source": "authored, fictional parties",
   "use_case": "Tax filing",
   "jurisdiction": "USF;DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/27-83b-election-deadline.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:28",
   "case_id": "28",
   "kind": "task",
   "title": "CCPA access request clock and extension",
   "source": "authored, fictional parties",
   "use_case": "Privacy request",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/28-ccpa-access-request-clock.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:29",
   "case_id": "29",
   "kind": "task",
   "title": "Delaware Section 220 demand triage",
   "source": "authored, fictional parties",
   "use_case": "Litigation triage",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/29-section-220-demand-triage.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:30",
   "case_id": "30",
   "kind": "task",
   "title": "Texas noncompete for a sales hire",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/30-texas-sales-noncompete.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:31",
   "case_id": "31",
   "kind": "task",
   "title": "Texas rent late fee refund",
   "source": "authored, fictional parties",
   "use_case": "Consumer dispute",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/31-texas-rent-late-fee.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:32",
   "case_id": "32",
   "kind": "task",
   "title": "Delaware franchise tax estimate recalculated",
   "source": "authored, fictional parties",
   "use_case": "Tax filing",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/32-delaware-franchise-tax-notice.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:33",
   "case_id": "33",
   "kind": "task",
   "title": "Event ticket resale: cancel and blacklist limits",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/33-event-ticket-resale-cancellation.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:34",
   "case_id": "34",
   "kind": "task",
   "title": "Two dismissals either side of 1 October 2026: tribunal deadlines",
   "source": "authored, fictional parties",
   "use_case": "Employment advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/34-unfair-dismissal-limit-straddles-1-oct-2026.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:35",
   "case_id": "35",
   "kind": "task",
   "title": "UAE supply contract breach after June 2026: which civil code",
   "source": "authored, fictional parties",
   "use_case": "Contract dispute",
   "jurisdiction": "AE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/35-uae-civil-transactions-law-2026.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:36",
   "case_id": "36",
   "kind": "task",
   "title": "FTC click-to-cancel status for US subscriptions",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "USF;CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/36-ftc-click-to-cancel-status.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:37",
   "case_id": "37",
   "kind": "task",
   "title": "EU AI Act timing for CV screening AI",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/37-eu-ai-act-omnibus-hr-screening.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:38",
   "case_id": "38",
   "kind": "task",
   "title": "Accredited investor test for a 506(b) angel",
   "source": "authored, fictional parties",
   "use_case": "Financing compliance",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/38-accredited-investor-ecfr-quote.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:39",
   "case_id": "39",
   "kind": "task",
   "title": "UK breach policy: 72 hours quote check",
   "source": "authored, fictional parties",
   "use_case": "Policy redline",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/39-uk-gdpr-breach-72-hours-quote-check.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:40",
   "case_id": "40",
   "kind": "task",
   "title": "Citation check on a draft letter with a suspect case",
   "source": "authored, fictional parties",
   "use_case": "Citation check",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/40-citation-check-brief-with-suspect-case.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:41",
   "case_id": "41",
   "kind": "task",
   "title": "Statutory late payment interest at the live base rate",
   "source": "authored, fictional parties",
   "use_case": "Debt recovery maths",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/41-late-payment-interest-live-base-rate.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:42",
   "case_id": "42",
   "kind": "task",
   "title": "California 2027 minimum wage and exempt salary",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/42-california-minimum-wage-and-exempt-salary-2027.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:43",
   "case_id": "43",
   "kind": "task",
   "title": "Companies House identity verification for a new director",
   "source": "authored, fictional parties",
   "use_case": "Corporate filing",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/43-companies-house-identity-verification-new-director.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:44",
   "case_id": "44",
   "kind": "task",
   "title": "VC fund 20% basket and exempt reporting adviser filing",
   "source": "authored, fictional parties",
   "use_case": "Regulatory compliance",
   "jurisdiction": "USF;CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/44-vc-fund-20-percent-basket-and-era-filing.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:45",
   "case_id": "45",
   "kind": "task",
   "title": "Stockholder written consent: delivery, notice, filing",
   "source": "authored, fictional parties",
   "use_case": "Corporate governance",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/45-delaware-stockholder-written-consent-notice.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:46",
   "case_id": "46",
   "kind": "task",
   "title": "Failure to prevent fraud: in scope through the group",
   "source": "authored, fictional parties",
   "use_case": "Regulatory advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/46-failure-to-prevent-fraud-group-scope.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:47",
   "case_id": "47",
   "kind": "task",
   "title": "Arbitration clause governing law after the 2025 Act",
   "source": "authored, fictional parties",
   "use_case": "Dispute strategy",
   "jurisdiction": "EW;NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/47-arbitration-agreement-governing-law-2025-act.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F1-mfn-waiver-docx-sign",
   "case_id": "FG-F1-mfn-waiver-docx-sign",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F1-mfn-waiver-docx-sign.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F10-tx-bounced-check",
   "case_id": "FG-F10-tx-bounced-check",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F10-tx-bounced-check.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F11-ew-no-fault-customer-derived",
   "case_id": "FG-F11-ew-no-fault-customer-derived",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F11-ew-no-fault-customer-derived.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F12-ca-sow-red-flags",
   "case_id": "FG-F12-ca-sow-red-flags",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F12-ca-sow-red-flags.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F13-mfn-mobile-continue-path",
   "case_id": "FG-F13-mfn-mobile-continue-path",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F13-mfn-mobile-continue-path.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F14-tx-write-it-now-date",
   "case_id": "FG-F14-tx-write-it-now-date",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F14-tx-write-it-now-date.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F15-safe-just-the-number",
   "case_id": "FG-F15-safe-just-the-number",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F15-safe-just-the-number.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F16-nda-no-questions-desktop",
   "case_id": "FG-F16-nda-no-questions-desktop",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F16-nda-no-questions-desktop.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F17-ew-glasgow-move-threat",
   "case_id": "FG-F17-ew-glasgow-move-threat",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F17-ew-glasgow-move-threat.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F18-ca-late-final-paycheck",
   "case_id": "FG-F18-ca-late-final-paycheck",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F18-ca-late-final-paycheck.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F19-clean-doc-investor-emails",
   "case_id": "FG-F19-clean-doc-investor-emails",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F19-clean-doc-investor-emails.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F2-safe-dilution",
   "case_id": "FG-F2-safe-dilution",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F2-safe-dilution.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F20-letter-then-email-it",
   "case_id": "FG-F20-letter-then-email-it",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F20-letter-then-email-it.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F21-ca-sign-tonight",
   "case_id": "FG-F21-ca-sign-tonight",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F21-ca-sign-tonight.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F22-nda-skip-approve-followup",
   "case_id": "FG-F22-nda-skip-approve-followup",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F22-nda-skip-approve-followup.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F23-safe-followup-math",
   "case_id": "FG-F23-safe-followup-math",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F23-safe-followup-math.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F24-ew-divorce-retest",
   "case_id": "FG-F24-ew-divorce-retest",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F24-ew-divorce-retest.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F25-redline-cover-email-docx",
   "case_id": "FG-F25-redline-cover-email-docx",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F25-redline-cover-email-docx.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F26-ca-group-layoff-52",
   "case_id": "FG-F26-ca-group-layoff-52",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F26-ca-group-layoff-52.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F27-docx-signing-email-garbage",
   "case_id": "FG-F27-docx-signing-email-garbage",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F27-docx-signing-email-garbage.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F28-nda-txt-control",
   "case_id": "FG-F28-nda-txt-control",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F28-nda-txt-control.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F29-docx-anon-vault-wall",
   "case_id": "FG-F29-docx-anon-vault-wall",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F29-docx-anon-vault-wall.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F3-texas-invoice",
   "case_id": "FG-F3-texas-invoice",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F3-texas-invoice.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F30-ca-employee-followup-approve",
   "case_id": "FG-F30-ca-employee-followup-approve",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F30-ca-employee-followup-approve.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F31-nda-txt-skip-followup-read",
   "case_id": "FG-F31-nda-txt-skip-followup-read",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F31-nda-txt-skip-followup-read.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F4-nda-review",
   "case_id": "FG-F4-nda-review",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F4-nda-review.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F5-england-divorce",
   "case_id": "FG-F5-england-divorce",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F5-england-divorce.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F6-ca-termination",
   "case_id": "FG-F6-ca-termination",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F6-ca-termination.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F7-angel-board-seat",
   "case_id": "FG-F7-angel-board-seat",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F7-angel-board-seat.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F8-series-a-one-shot-mfn-sign",
   "case_id": "FG-F8-series-a-one-shot-mfn-sign",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F8-series-a-one-shot-mfn-sign.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-F9-ca-severance-age-44",
   "case_id": "FG-F9-ca-severance-age-44",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-F9-ca-severance-age-44.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L1-bank-counterparty-dora",
   "case_id": "FG-L1-bank-counterparty-dora",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L1-bank-counterparty-dora.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L10-eis-connected-persons",
   "case_id": "FG-L10-eis-connected-persons",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L10-eis-connected-persons.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L11-bank-breach-shared-position",
   "case_id": "FG-L11-bank-breach-shared-position",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L11-bank-breach-shared-position.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L12-uk-financial-promotion",
   "case_id": "FG-L12-uk-financial-promotion",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L12-uk-financial-promotion.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L13-controller-before-dpa",
   "case_id": "FG-L13-controller-before-dpa",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L13-controller-before-dpa.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L14-penalty-clause-check",
   "case_id": "FG-L14-penalty-clause-check",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L14-penalty-clause-check.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L15-ca-work-for-hire-cite",
   "case_id": "FG-L15-ca-work-for-hire-cite",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L15-ca-work-for-hire-cite.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L16-nsi-threshold-cite",
   "case_id": "FG-L16-nsi-threshold-cite",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L16-nsi-threshold-cite.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L17-uk-board-resolution-sources",
   "case_id": "FG-L17-uk-board-resolution-sources",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L17-uk-board-resolution-sources.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L18-uae-freelancer-services",
   "case_id": "FG-L18-uae-freelancer-services",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L18-uae-freelancer-services.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L2-answer-first-intake",
   "case_id": "FG-L2-answer-first-intake",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L2-answer-first-intake.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L3-lawyer-quote-consent",
   "case_id": "FG-L3-lawyer-quote-consent",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L3-lawyer-quote-consent.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L4-first-view-latency",
   "case_id": "FG-L4-first-view-latency",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L4-first-view-latency.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L5-governing-law-transfer",
   "case_id": "FG-L5-governing-law-transfer",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L5-governing-law-transfer.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L6-bank-breach-notice",
   "case_id": "FG-L6-bank-breach-notice",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L6-bank-breach-notice.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L7-critical-function-art30",
   "case_id": "FG-L7-critical-function-art30",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L7-critical-function-art30.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L8-privacy-shield-dpf",
   "case_id": "FG-L8-privacy-shield-dpf",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L8-privacy-shield-dpf.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:FG-L9-review-label-honesty",
   "case_id": "FG-L9-review-label-honesty",
   "kind": "task",
   "title": null,
   "source": "focus group, anonymised real inputs",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/FG-L9-review-label-honesty.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:R4",
   "case_id": "R4",
   "kind": "task",
   "title": "Regression: internal text leaks",
   "source": "authored, fictional parties",
   "use_case": "Regression",
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/R4-regression-internal-text-leaks.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-01",
   "case_id": "RT-01",
   "kind": "task",
   "title": "MFN waiver from a messy uploaded SAFE, with the Strategic Safe carve-out trap",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-01-mfn-waiver-messy-safe-strategic-trap.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-02",
   "case_id": "RT-02",
   "kind": "task",
   "title": "Mid-flow correction: 'no I meant the MFN waiver', then send",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 3.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-02-side-letter-correction-to-mfn-waiver-then-send.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-03",
   "case_id": "RT-03",
   "kind": "task",
   "title": "Board consent: strategic SAFE plus three option grants at the 409A price, sent to both directors",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-03-board-consent-strategic-safe-and-option-grants.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-04",
   "case_id": "RT-04",
   "kind": "task",
   "title": "Employee early exercise: 83(b) clock, who files, reminder",
   "source": "authored, fictional parties",
   "use_case": "Tax filing",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-04-employee-early-exercise-83b.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-05",
   "case_id": "RT-05",
   "kind": "task",
   "title": "'Pay $425 or lose good standing' letter vs the real Delaware annual report and tax",
   "source": "authored, fictional parties",
   "use_case": "Tax filing",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-05-delaware-solicitation-letter-and-real-tax.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-06",
   "case_id": "RT-06",
   "kind": "task",
   "title": "Departing cofounder: IP he says is 'his', and unvested shares back (California)",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "CA;DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-06-departing-cofounder-ip-and-unvested-shares.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-07",
   "case_id": "RT-07",
   "kind": "task",
   "title": "Advisor asks for 1% plus $5k a month: counter, then fill the Vault template",
   "source": "authored, fictional parties",
   "use_case": "Drafting + negotiation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-07-advisor-counter-and-template-fill.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-08",
   "case_id": "RT-08",
   "kind": "task",
   "title": "SAFE round close-out: signed, wired, and one wire with no signed SAFE",
   "source": "authored, fictional parties",
   "use_case": "Financing ops",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-08-safe-round-close-out-reconciliation.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-09",
   "case_id": "RT-09",
   "kind": "task",
   "title": "Side letter with an observer seat and explicit no MFN, no pro rata",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-09-side-letter-observer-no-mfn-no-pro-rata.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-10",
   "case_id": "RT-10",
   "kind": "task",
   "title": "Exploding seed term sheet: founder ownership if we accept vs counter at a 15% pool",
   "source": "authored, fictional parties",
   "use_case": "Financing maths",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-10-seed-term-sheet-pool-shuffle-math.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-11",
   "case_id": "RT-11",
   "kind": "task",
   "title": "Map a VC's DD list against the Vault, and catch that the Vault is the wrong company",
   "source": "authored, fictional parties",
   "use_case": "Diligence",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-11-dd-list-vs-vault-company-mismatch.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-12",
   "case_id": "RT-12",
   "kind": "task",
   "title": "Angel asks for pro rata: yes, next priced round only, clause plus email",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-12-pro-rata-yes-next-priced-round-only.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-13",
   "case_id": "RT-13",
   "kind": "task",
   "title": "Pool increase of 500,000: board and stockholder consents, and an authorized-shares gap",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-13-option-pool-increase-authorized-shares-check.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-14",
   "case_id": "RT-14",
   "kind": "task",
   "title": "Clean up a SAFE's defined terms without touching the economics",
   "source": "authored, fictional parties",
   "use_case": "Edit own document",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-14-safe-cleanup-defined-terms-no-economics.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-15",
   "case_id": "RT-15",
   "kind": "task",
   "title": "VC side: draft a seed term sheet and send it to the founder",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-15-vc-drafts-seed-term-sheet-sends-to-founder.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-16",
   "case_id": "RT-16",
   "kind": "task",
   "title": "VC confirmatory DD checklist for a UK fintech raising on SEIS and EIS",
   "source": "authored, fictional parties",
   "use_case": "Diligence",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-16-vc-dd-checklist-uk-fintech-seis-eis.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-17",
   "case_id": "RT-17",
   "kind": "task",
   "title": "LP side letter for a family office into a Cayman fund",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-17-lp-side-letter-cayman-fund.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-18",
   "case_id": "RT-18",
   "kind": "task",
   "title": "Investor side: should we elect the new SAFE terms under our MFN?",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-18-investor-mfn-election-decision.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-19",
   "case_id": "RT-19",
   "kind": "task",
   "title": "Customer-side redline of vendor paper: 36-month auto-renew, data training, posted terms prevail",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-19-vendor-msa-customer-side-redline-eod.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-20",
   "case_id": "RT-20",
   "kind": "task",
   "title": "Open S3 bucket with EU customer emails: the 72-hour clock across the DST change",
   "source": "authored, fictional parties",
   "use_case": "Privacy incident",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-20-eu-breach-72-hours-across-dst.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-21",
   "case_id": "RT-21",
   "kind": "task",
   "title": "Layoff of 60 at the SF office: WARN, Cal-WARN, OWBPA group disclosures and final pay",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-21-us-layoff-warn-cal-warn-owbpa-plan.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-22",
   "case_id": "RT-22",
   "kind": "task",
   "title": "UK restructure: 24 roles in London, CEO wants everyone out in three weeks",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-22-uk-collective-redundancy-timeline.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-23",
   "case_id": "RT-23",
   "kind": "task",
   "title": "Separation agreement for a 52-year-old, then HR asks to cut 21 days to 3",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "CA;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-23-separation-agreement-52-hr-pushback.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-24",
   "case_id": "RT-24",
   "kind": "task",
   "title": "Hiring a designer in Lisbon with no EU entity: employee, contractor or EOR",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "EU;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-24-lisbon-hire-no-eu-entity.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-25",
   "case_id": "RT-25",
   "kind": "task",
   "title": "Privacy policy for a health scheduling app using an AI subprocessor",
   "source": "authored, fictional parties",
   "use_case": "Privacy drafting",
   "jurisdiction": "USF;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-25-privacy-policy-health-app-ai-subprocessor.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-26",
   "case_id": "RT-26",
   "kind": "task",
   "title": "Customer wants unlimited data breach liability: counter with a 2x super-cap",
   "source": "authored, fictional parties",
   "use_case": "Contract negotiation",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-26-unlimited-breach-liability-super-cap-counter.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-27",
   "case_id": "RT-27",
   "kind": "task",
   "title": "German customer's AV-Vertrag: what it is and what to change",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "EU",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-27-german-av-vertrag-is-a-dpa.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-28",
   "case_id": "RT-28",
   "kind": "task",
   "title": "Draft our enterprise MSA and save it as a Vault template",
   "source": "authored, fictional parties",
   "use_case": "Template creation",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-28-enterprise-msa-template-saved-to-vault.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-29",
   "case_id": "RT-29",
   "kind": "task",
   "title": "Directors' written resolution approving a GBP 3m sponsorship, sent to 3 directors",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-29-uk-directors-written-resolution-sponsorship.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-30",
   "case_id": "RT-30",
   "kind": "task",
   "title": "What do we refund if only 3 races run? (and is the term sheet even binding)",
   "source": "authored, fictional parties",
   "use_case": "Contract maths",
   "jurisdiction": "EW;CH",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-30-sponsorship-make-good-math-non-binding-catch.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-31",
   "case_id": "RT-31",
   "kind": "task",
   "title": "Sponsorship negotiation: counsel's reply pasted, draft the counter and the email",
   "source": "authored, fictional parties",
   "use_case": "Contract negotiation",
   "jurisdiction": "EW;CH",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-31-sponsorship-counter-after-counsel-reply.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-32",
   "case_id": "RT-32",
   "kind": "task",
   "title": "Driver wants a crypto exchange as personal sponsor: is crypto 'betting'?",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "CH;EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-32-driver-personal-sponsor-crypto-category.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-33",
   "case_id": "RT-33",
   "kind": "task",
   "title": "Driver did 31 commercial days: what we owe and the note to her manager",
   "source": "authored, fictional parties",
   "use_case": "Contract maths",
   "jurisdiction": "CH",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-33-driver-commercial-days-overage.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-34",
   "case_id": "RT-34",
   "kind": "task",
   "title": "Ticket T&Cs go live Monday: UK consumer problems and a redline",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-34-ticket-terms-pre-launch-uk.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-35",
   "case_id": "RT-35",
   "kind": "task",
   "title": "Abu Dhabi onshore venue hosting agreement: promoter's markup",
   "source": "authored, fictional parties",
   "use_case": "Contract redline",
   "jurisdiction": "AE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-35-abu-dhabi-onshore-venue-hosting-markup.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-36",
   "case_id": "RT-36",
   "kind": "task",
   "title": "Official supplier agreement: drinks for three festivals, value in kind plus cash",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-36-official-supplier-agreement-value-in-kind.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-37",
   "case_id": "RT-37",
   "kind": "task",
   "title": "Highlights licence to a streamer, with a conflict hidden in the venue deal",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "EW;AE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-37-highlights-licence-gulf-rights-conflict.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-38",
   "case_id": "RT-38",
   "kind": "task",
   "title": "Customer refuses to pay over a fixed defect: firm but friendly letter",
   "source": "authored, fictional parties",
   "use_case": "Debt recovery",
   "jurisdiction": "TX",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-38-disputed-invoice-firm-not-nuclear-texas.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-39",
   "case_id": "RT-39",
   "kind": "task",
   "title": "Bakery wants out at the break date, but the notice window has passed",
   "source": "authored, fictional parties",
   "use_case": "Property",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-39-lease-break-notice-deadline-passed.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-40",
   "case_id": "RT-40",
   "kind": "task",
   "title": "One-page cleaning services agreement, sent to the client",
   "source": "authored, fictional parties",
   "use_case": "Drafting + e-sign",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-40-cleaning-services-one-pager-arizona.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-41",
   "case_id": "RT-41",
   "kind": "task",
   "title": "Executor in England: mother died in August, house in her sole name",
   "source": "authored, fictional parties",
   "use_case": "Personal legal advice",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-41-england-probate-executor-sole-name-house.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-42",
   "case_id": "RT-42",
   "kind": "task",
   "title": "Two-member LLC: partner wants out, what document and a draft",
   "source": "authored, fictional parties",
   "use_case": "Drafting",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-42-two-person-llc-partner-buyout-oregon.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-43",
   "case_id": "RT-43",
   "kind": "task",
   "title": "Employer in Fresno: can the last check go out with Friday payroll?",
   "source": "authored, fictional parties",
   "use_case": "Employment",
   "jurisdiction": "CA",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-43-fresno-employer-final-pay-timing.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-44",
   "case_id": "RT-44",
   "kind": "task",
   "title": "Draft the non-renewal notice now and remind me before the window",
   "source": "authored, fictional parties",
   "use_case": "Obligations",
   "jurisdiction": "NY",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-44-non-renewal-notice-and-reminder.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-45",
   "case_id": "RT-45",
   "kind": "task",
   "title": "Signer side: request changes on an order form instead of signing",
   "source": "authored, fictional parties",
   "use_case": "Signing flow (recipient)",
   "jurisdiction": "EW;USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-45-recipient-request-changes-order-form.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-46",
   "case_id": "RT-46",
   "kind": "task",
   "title": "'Just sign it for me, I trust them'",
   "source": "authored, fictional parties",
   "use_case": "Signing flow (recipient)",
   "jurisdiction": "USF",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-46-recipient-just-sign-it-for-me.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-47",
   "case_id": "RT-47",
   "kind": "task",
   "title": "Investor sends back the waiver with her own edits: accept, counter, resend",
   "source": "authored, fictional parties",
   "use_case": "Signing flow (negotiate)",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-47-investor-returned-waiver-with-edits.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-48",
   "case_id": "RT-48",
   "kind": "task",
   "title": "Onshore UAE contract still cites the 1985 Civil Code: amend it",
   "source": "authored, fictional parties",
   "use_case": "Edit own document + e-sign",
   "jurisdiction": "AE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 1.5,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-48-uae-onshore-amend-repealed-civil-code-reference.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-49",
   "case_id": "RT-49",
   "kind": "task",
   "title": "NDA with hidden instructions to the AI: review it anyway, safely",
   "source": "authored, fictional parties",
   "use_case": "Contract review",
   "jurisdiction": "EW",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-49-nda-hidden-prompt-injection.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:RT-50",
   "case_id": "RT-50",
   "kind": "task",
   "title": "'Backdate the board consent to Sept 1': refuse, offer ratification",
   "source": "authored, fictional parties",
   "use_case": "Behavior",
   "jurisdiction": "DE",
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": 2.0,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/RT-50-backdate-request-ratification-instead.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-B1-vc-board-resolution",
   "case_id": "UC-B1-vc-board-resolution",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-B1-vc-board-resolution.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-B2-vc-safe-vs-standard",
   "case_id": "UC-B2-vc-safe-vs-standard",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-B2-vc-safe-vs-standard.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-B3-vc-term-sheet-check",
   "case_id": "UC-B3-vc-term-sheet-check",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-B3-vc-term-sheet-check.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-E1-editor-make-mutual",
   "case_id": "UC-E1-editor-make-mutual",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-E1-editor-make-mutual.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-E2-editor-tighten-indemnity",
   "case_id": "UC-E2-editor-tighten-indemnity",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-E2-editor-tighten-indemnity.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-E3-editor-response-to-comments",
   "case_id": "UC-E3-editor-response-to-comments",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-E3-editor-response-to-comments.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-M1-markup-dpa-response",
   "case_id": "UC-M1-markup-dpa-response",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-M1-markup-dpa-response.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-M2-markup-nda-vendor",
   "case_id": "UC-M2-markup-nda-vendor",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-M2-markup-nda-vendor.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-O1-onboarding-first-session",
   "case_id": "UC-O1-onboarding-first-session",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-O1-onboarding-first-session.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-P1-payments-msa-schedule",
   "case_id": "UC-P1-payments-msa-schedule",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-P1-payments-msa-schedule.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-P2-payments-safe-milestones",
   "case_id": "UC-P2-payments-safe-milestones",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-P2-payments-safe-milestones.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-S1-sim-liability-cap",
   "case_id": "UC-S1-sim-liability-cap",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-S1-sim-liability-cap.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-S2-sim-safe-vs-priced",
   "case_id": "UC-S2-sim-safe-vs-priced",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-S2-sim-safe-vs-priced.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-S3-sim-lease-break",
   "case_id": "UC-S3-sim-lease-break",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-S3-sim-lease-break.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-V1-vault-upload-summaries",
   "case_id": "UC-V1-vault-upload-summaries",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-V1-vault-upload-summaries.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-V2-vault-renewals-90-days",
   "case_id": "UC-V2-vault-renewals-90-days",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-V2-vault-renewals-90-days.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  },
  {
   "row_id": "task:UC-V3-vault-personal-lease-insurance",
   "case_id": "UC-V3-vault-personal-lease-insurance",
   "kind": "task",
   "title": null,
   "source": "authored, fictional parties",
   "use_case": null,
   "jurisdiction": null,
   "customer": null,
   "doc_type": null,
   "model": null,
   "prompt_version": null,
   "run": null,
   "split": null,
   "sample": null,
   "scale": null,
   "accuracy": null,
   "grounding": null,
   "specificity": null,
   "actionability": null,
   "polish": null,
   "e2e_wow": null,
   "task_score": null,
   "weight": null,
   "hard_fail": null,
   "metric": null,
   "metric_value": null,
   "basis": "not_run",
   "label": "NOT RUN",
   "judge": null,
   "run_date": null,
   "cost_usd": null,
   "latency_s": null,
   "link": "evals/tasks/UC-V3-vault-personal-lease-insurance.md",
   "notes": "Task file on disk; not yet run.",
   "segment": null
  }
 ]
}