outputs/official_release_audit.json

{
  "all_passed": true,
  "claim1_math": {
    "coverage_shortfall_percentage_points": 0.4545454545454519,
    "dcf_coverage": 0.9654545454545455,
    "dcf_retained": 1.7604545454545453,
    "frequency_baseline_retained": 0.7255445544554455,
    "literal_verdict": "falsified_as_composite_reliability_claim",
    "relative_improvement_percent": 142.63906808257244,
    "target_coverage": 0.97
  },
  "claim2_felm": {
    "dcf_coverage": 0.9915476190476191,
    "dcf_retained": 0.7153174603174602,
    "frequency_baseline_retained": 0.4446760563380282,
    "relative_improvement_percent": 60.86259876643758,
    "target_coverage": 0.99,
    "target_met": true
  },
  "claim3_calibration": {
    "native_actual_graph_nonconformity_evaluations": 50,
    "released_convergence_suites": 5,
    "released_trials_per_suite": 20,
    "theorem_scope": "soft nonconformity-score convergence; quantile recovery is a separate source statement"
  },
  "claim4_prediction": {
    "native_actual_graph_prediction_nodes": 503,
    "released_convergence_suites": 2,
    "released_trials_per_suite": 20
  },
  "claim5_agreement": {
    "all_between_90_and_100_percent": true,
    "cv_folds": 20,
    "maximum_agreement": 1.0,
    "minimum_agreement": 0.9021917808219178,
    "predictions_per_row": 14600,
    "rows": 10
  },
  "claim6_joint_pipeline": {
    "actual_claim_nodes": 503,
    "actual_graphs": 50,
    "ancestor_removal_l1_difference": 26.071823805570602,
    "end_to_end_gradient_finite_and_nonzero": true
  },
  "official_commit": "0b4d5487a9868a18c4f9aa5b3d96cdccc705ca97",
  "paper_id": "XfndtVLIub"
}