{
  "release_date": "2026-10-06",
  "dataset": "MaYiding/OracleProto",
  "split": "test",
  "rows": 300,
  "full_points": 528,
  "sampling_n": 3,
  "complete_configurations": 27,
  "resolved_date_range": [
    "2026-03-12",
    "2026-09-21"
  ],
  "type_counts": {
    "yes_no": 100,
    "binary_named": 9,
    "mc_single": 154,
    "mc_multi": 37
  },
  "source_database_sha256": "67243aeadbab783202d0c715c5dd01cef265ab41bf729248ff0239d6bcb06c97",
  "csv_sha256": "b5fa948d8c29044823dbc531065390543715b451918da1efb87d90f110c1c6b0",
  "prompt_recipe_sha256": "7096361043a30bc18d7ed25701ec01caba12483d7dac564754f82896218e94b8",
  "scoring_contract": {
    "aggregation": "100*sum(points*mean_trial_credit)/sum(points)",
    "bootstrap": "paired_question_clusters_stratified_by_question_type",
    "bootstrap_iterations": 2000,
    "bootstrap_seed": 20261001,
    "ci_level": 0.95,
    "content_policy": "zero_credit_in_denominator",
    "cutoff": "exclude_question",
    "diagnostics": "question_mean_then_population_mean",
    "headline_population": "intersection_of_declared_admissible_questions",
    "invalid_answer": "zero_credit_in_denominator",
    "missing_or_infrastructure_error": "incomplete_no_official_score",
    "multi_credit": "2TP/(2TP+FP+FN)",
    "name": "points-f1",
    "pass_at_k": "mean(1-comb(K-c,k)/comb(K,k))_exact_match",
    "points": {
      "binary_named": 1,
      "mc_multi": 3,
      "mc_single": 2,
      "yes_no": 1
    },
    "single_credit": "exact_set_equality",
    "vote": "strict_majority_of_complete_answer_sets_invalid_votes_abstain"
  },
  "cost_basis": "Public standard online tariffs repriced per request, sum of three repetitions divided by three; forecasting-model inference only. Excludes search, leakage-detector usage, missing usage, and invoice totals.",
  "leakage_audit": {
    "source": "author-reported final audit",
    "auditor": "GPT 6.1 Sol",
    "reviewed": 10000,
    "positive_labels": 137,
    "rate_percent": 1.37,
    "human_verified_positives": 137,
    "human_confirmed_leaks": 137,
    "human_verified_negative_labels": 1000,
    "human_found_leaks_in_negative_sample": 0
  },
  "results_source": "Frozen complete 300-question panel"
}