{
  "review_date": "2026-10-03",
  "scope": "Editorial reconstruction of completed labeling and adjudication; no models or experiments rerun.",
  "official_submission_query_date": "2026-10-02",
  "submissions": [
    {
      "version": 83,
      "submission_ref": 55967504,
      "description": "modern-universe-verifier-refit-exact-v83-shaff9af4c0 (p/dv/v4 policy v12 bindings v7 thr0.75 cap50)",
      "Public": "0.94437",
      "Private": "0.91331"
    },
    {
      "version": 85,
      "submission_ref": 56022669,
      "description": "hand-label-verifier-exact-v85-sha00062d88",
      "Public": "0.93996",
      "Private": "0.91373"
    },
    {
      "version": 86,
      "submission_ref": 56030957,
      "description": "batch2-hand-verifier-precision-frontier-exact-v86-sha23b39ba1",
      "Public": "0.94347",
      "Private": "0.90877"
    },
    {
      "version": 87,
      "submission_ref": 56038152,
      "description": "runtime-v2-features-batch2-0.90-exact-v87-shae07add44 (probe vs v86)",
      "Public": "0.94662",
      "Private": "0.91217"
    }
  ],
  "comparisons": [
    {
      "comparator": 83,
      "candidate": 85,
      "Public_delta": "-0.00441",
      "Private_delta": "0.00042"
    },
    {
      "comparator": 83,
      "candidate": 86,
      "Public_delta": "-0.00090",
      "Private_delta": "-0.00454"
    },
    {
      "comparator": 86,
      "candidate": 87,
      "Public_delta": "0.00315",
      "Private_delta": "0.00340"
    },
    {
      "comparator": 83,
      "candidate": 87,
      "Public_delta": "0.00225",
      "Private_delta": "-0.00114"
    }
  ],
  "selection": {
    "batch1": {
      "items": 500,
      "blind_gold": 50,
      "minimum_score": 0.5
    },
    "batch2": {
      "items": 500,
      "blind_gold": 50,
      "minimum_score": 0.4
    },
    "presentation": "Scores hidden; t-1,t,t+1,t+2 XY z-slab max projections, zoom and an x-z side view at t+1."
  },
  "actual_ingestion": {
    "v85": {
      "labeled": 250,
      "y": 66,
      "n": 86,
      "s": 98,
      "gold_judged": 16,
      "gold_tp": 7,
      "gold_fp": 1,
      "gold_fn": 1,
      "gold_tn": 7,
      "labeler_precision_on_gold": 0.875,
      "labeler_recall_on_gold": 0.875,
      "labeler_accuracy_on_gold": 0.875,
      "new_rows": 136,
      "new_positives": 58,
      "base_rows": 527448,
      "base_positives": 174
    },
    "v86_and_v87": {
      "labeled": 500,
      "y": 97,
      "n": 207,
      "s": 196,
      "gold_judged": 39,
      "gold_tp": 18,
      "gold_fp": 0,
      "gold_fn": 4,
      "gold_tn": 17,
      "labeler_precision_on_gold": 1.0,
      "labeler_recall_on_gold": 0.818,
      "labeler_accuracy_on_gold": 0.897,
      "new_rows": 265,
      "new_positives": 79,
      "base_rows": 527448,
      "base_positives": 174
    }
  },
  "final_review": {
    "batches": [
      {
        "batch": 1,
        "path": "results/hand_labels/batch1/hand_labels_batch1_final.json",
        "counts": {
          "n": 156,
          "s": 250,
          "y": 94
        }
      },
      {
        "batch": 2,
        "path": "results/hand_labels/batch2/hand_labels_batch2_final.json",
        "counts": {
          "s": 236,
          "n": 167,
          "y": 97
        }
      }
    ],
    "unknown_total": 486,
    "non_gold_n_to_s_changes": 60,
    "gold_gt_positive_labeled_negative": 7,
    "user_correction": "Negative judgments were not distance-driven; depth-ambiguous cases were moved to unknown. The distance-heuristic account is a historical hypothesis, not a confirmed cause."
  },
  "local_scope": "Verifier fit on one embryo prefix including its hand rows and applied to the other over 199 movies with fixed upstream replay assets; reused development evidence, not complete independently selected whole-pipeline OOF.",
  "local_v85": {
    "universe": "modern_oof_v1 a000 replay (deployed stack, twofold e200 pair, embryo-out), fork-free base 0.7489 (44b6 0.8095 / 6bba 0.7383)",
    "protocol": "kernel-faithful: the SHIPPED runtime module (scripts/run_runtime_verifier_oof.py) fit prefix-pure (train one embryo prefix incl. its _hand rows, apply the other), official scorer, 199 movies; cap 50",
    "table": "v83 union table (527,448 rows, 174 positives) + labeling batch 1 items 1-250 hand-judged (y 66 / n 86 / s 98; gold agreement 14/16; 136 non-gold rows ingested: 58 positives, 78 negatives) = 232 positives",
    "deployed_v83_table_same_protocol": {
      "0.75": {
        "overall": 0.7535,
        "44b6": 0.8161,
        "6bba": 0.7425,
        "div_tp": 8,
        "div_fp": 19
      }
    },
    "hand_label_table_sweep": {
      "0.65": {
        "overall": 0.7583,
        "44b6": 0.8206,
        "6bba": 0.7475,
        "div_tp": 24,
        "div_fp": 96
      },
      "0.70": {
        "overall": 0.7575,
        "44b6": 0.8184,
        "6bba": 0.7469,
        "div_tp": 21,
        "div_fp": 86
      },
      "0.75": {
        "overall": 0.7573,
        "44b6": 0.8163,
        "6bba": 0.7469,
        "div_tp": 19,
        "div_fp": 71
      },
      "0.85": {
        "overall": 0.755,
        "44b6": 0.8169,
        "6bba": 0.7442,
        "div_tp": 12,
        "div_fp": 43
      },
      "alltrain_in_sample_0.75": {
        "overall": 0.7646,
        "44b6": 0.8279,
        "6bba": 0.7535,
        "div_tp": 40,
        "div_fp": 100
      }
    },
    "chosen_operating_point": {
      "threshold": 0.65,
      "max_per_movie": 50,
      "delta_vs_deployed_table": 0.0048,
      "44b6_delta": 0.0045,
      "6bba_delta": 0.005,
      "rule": "predeclared: >= +0.001 over the deployed point with both prefixes non-negative; monotone 0.65 > 0.70 > 0.75 > 0.85"
    },
    "label_learning_curve": "random GT positives: 43 -> 0.7495, 86 -> 0.7510, 128 -> 0.7515, 174 -> 0.7535 (+0.003 per 100, linear); boundary-band hand positives: +58 -> +0.0038 at 0.75",
    "hidden_context": "v83 (this verifier family at 0.75, 174 positives) = public 0.944, local +0.0065 -> public +0.007 (transfer ~1.1x)",
    "measured": "2026-09-05, pod:results/division_p1/{handlabel_price_partial250,label_curve_v1}"
  },
  "local_v87": {
    "universe": "modern_oof_v1 a000 replay (deployed stack, twofold e200 pair, embryo-out), fork-free base 0.7489 (44b6 0.8095 / 6bba 0.7383)",
    "protocol": "kernel-faithful: the SHIPPED runtime module (v2, sha d16d5686) fit prefix-pure (train one embryo prefix incl. its _hand rows, apply the other), official scorer, 199 movies; cap 50",
    "runtime_v2": "28 features = the 25 of v1 + the labeler's cue: q_t_node_um (nearest detection at t to the candidate daughter, excl. the parent and its predecessor), q_t_bright_rel / q_tm1_bright_rel (appearance z-score at the daughter's position sampled at t / t-1 minus at t+1). Absent values (old-universe training rows) are median-imputed at fit; NaN routing cost 8 TP in the in-table ablation.",
    "table": "v83 union table (527,448 rows, 174 positives) + labeling BATCH 2 (79 hand positives, 186 negatives; gold precision 1.00) = 253 positives; 284,555 modern/hand rows carry the v2 columns",
    "deployed_v83_table_v1_runtime": {
      "0.75": {
        "overall": 0.7535,
        "44b6": 0.8161,
        "6bba": 0.7425,
        "div_tp": 8,
        "div_fp": 19,
        "picks_199_movies": 2242,
        "expected_pick_precision": 0.41
      }
    },
    "v86_v1_runtime_batch2_table": {
      "0.90": {
        "overall": 0.7591,
        "44b6": 0.8167,
        "6bba": 0.7491,
        "div_tp": 18,
        "div_fp": 25,
        "picks_199_movies": 2324,
        "expected_pick_precision": 0.825
      }
    },
    "v2_runtime_batch2_table_sweep": {
      "0.95": {
        "overall": 0.7569,
        "44b6": 0.8131,
        "6bba": 0.7472,
        "div_tp": 14,
        "div_fp": 23,
        "picks_199_movies": 1148,
        "expected_pick_precision": 0.85
      },
      "0.90": {
        "overall": 0.7609,
        "44b6": 0.8165,
        "6bba": 0.7513,
        "div_tp": 23,
        "div_fp": 40,
        "picks_199_movies": 2704,
        "expected_pick_precision": 0.821
      },
      "0.85": {
        "overall": 0.7612,
        "44b6": 0.816,
        "6bba": 0.7517,
        "div_tp": 25,
        "div_fp": 51,
        "picks_199_movies": 3880,
        "expected_pick_precision": 0.7
      },
      "0.75": {
        "overall": 0.761,
        "44b6": 0.823,
        "6bba": 0.7502,
        "div_tp": 29,
        "div_fp": 84
      }
    },
    "control_v83_table_v2_runtime": {
      "0.75": {
        "overall": 0.7536,
        "div_tp": 8,
        "div_fp": 18
      },
      "0.90": {
        "overall": 0.7489,
        "div_tp": 0,
        "div_fp": 3
      },
      "note": "v2 features add nothing without the hand labels"
    },
    "hand_row_oof_precision_by_band_v2": {
      "[0.95,1]": 0.85,
      "[0.90,0.95)": 0.8,
      "[0.85,0.90)": 0.42,
      "[0.75,0.85)": 0.29,
      "[0.65,0.75)": 0.12,
      "cumulative_>=0.90": 0.83
    },
    "chosen_operating_point": {
      "threshold": 0.9,
      "max_per_movie": 50,
      "local_delta_vs_v83_table": 0.0074,
      "44b6_delta": 0.0004,
      "6bba_delta": 0.0088,
      "rule": "precision frontier: expected pick precision 0.82 (= v86) with 16% more picks (2,704 vs 2,324 over 199 movies); v85 (0.65, precision ~0.5) lost 0.005 public"
    },
    "measured": "2026-09-05, pod:results/division_p1/handlabel_price_{b2_v2i,v83_v2i}/chain.log; scripts/{hand_band_precision,pick_population_precision,v2_feature_ablation_quick}.py"
  },
  "missed_gt_6bba": {
    "held_out_gt_events_with_candidate": 126,
    "found": 25,
    "missed": 101,
    "missed_score_pct": [
      0.023,
      0.084,
      0.368
    ]
  },
  "same_recipe_more_labels": {
    "batch2_v2_thr090": {
      "score": 0.7609,
      "TP": 23,
      "FP": 40
    },
    "cleaned_batch1_plus_batch2_v2_thr090": {
      "score": 0.7602,
      "TP": 25,
      "FP": 67
    }
  },
  "equal_fp_example": {
    "v1_thr085": {
      "TP": 22,
      "FP": 40
    },
    "v2_thr090": {
      "TP": 23,
      "FP": 40
    }
  },
  "S4": {
    "purpose": "Blinded crossing-swap diagnostic, never training data.",
    "label_counts": {
      "a": 28,
      "b": 22,
      "u": 48,
      "x": 2
    },
    "controls": {
      "agreement": {
        "ci90": [
          0.5444175959982525,
          0.8959191640898638
        ],
        "k": 15,
        "n": 20,
        "share": 0.75
      },
      "counts": {
        "gt": 15,
        "missing": 0,
        "other": 0,
        "u": 5,
        "x": 0
      },
      "missing": 0,
      "n": 20,
      "neither_x": {
        "ci90": [
          0.0,
          0.13910834066826522
        ],
        "k": 0,
        "n": 20,
        "share": 0.0
      },
      "unsure_or_neither": {
        "ci90": [
          0.10408083591013584,
          0.455582404001749
        ],
        "k": 5,
        "n": 20,
        "share": 0.25
      },
      "unsure_u": {
        "ci90": [
          0.10408083591013584,
          0.455582404001749
        ],
        "k": 5,
        "n": 20,
        "share": 0.25
      },
      "wrong_letter": {
        "ci90": [
          0.0,
          0.13910834066826522
        ],
        "k": 0,
        "n": 20,
        "share": 0.0
      }
    },
    "rule": {
      "applied_to": "point estimates",
      "ceiling": "pooled gt_supported < 0.35",
      "gt_noise_material": "pooled prediction_supported >= 0.15",
      "labeler_trusted": "control_agreement >= 0.85",
      "mixed": "otherwise",
      "resolvable": "pooled gt_supported >= 0.5"
    },
    "verdict": {
      "gt_noise_material": null,
      "labeler_trusted": false,
      "note": "control_agreement < 0.85: the batch is inconclusive and the swap shares are not read",
      "status": "complete",
      "swap_class": null,
      "verdict": "inconclusive"
    },
    "not_done": "No labels enter any table, model or threshold. No second batch unless inconclusive, and then only by user decision."
  },
  "limits": [
    "Private differences are observed whole-verifier-package results, not per-cue causal estimates; Private error labels are unavailable.",
    "v83 to v85 and v85 to v86 change both labels and threshold; v86 to v87 keeps judgments, row composition, threshold and upstream models fixed while adding features, imputation and refit.",
    "The final review JSONs are not the label set deployed in v87.",
    "Label noise, population conflict and classifier capacity were not experimentally separated.",
    "A failed promotion gate does not exhaust annotation, alternative representations or existing-candidate learning.",
    "Unrun successor experiments do not establish a counterfactual medal."
  ],
  "sources": [
    {
      "path": "_draft/postcompetition-review-20261002/references/own-submissions-20261002.json",
      "sha256": "927dfd7b60ae05dd39b20c57f9c49dd1dd9830cc7ff04aa6ea19d3ef494408db"
    },
    {
      "path": "_draft/postcompetition-review-20261002/references/own-outcome-audit-20261002.json",
      "sha256": "a61d839c4957328dfb612ccad714f2a9863fcf88becc95270d5a7d911648f793"
    },
    {
      "path": "scripts/select_labeling_batch.py",
      "sha256": "1e1f2c09ea15e9f6d4792e5dbc6fad0eed026a1f981cd867e8b526c658ccd74b"
    },
    {
      "path": "scripts/ingest_hand_labels.py",
      "sha256": "d9e388fd30c40e436b71698166bb202cecf5599f9a91a5a95be454dbf80ebee3"
    },
    {
      "path": "scripts/render_labeling_crops.py",
      "sha256": "5ef597a781bb341a100ef367102ee1dbe3579df1d8bc3d6761a829778362df0d"
    },
    {
      "path": "results/hand_labels/batch1/batch1.csv",
      "sha256": "c25109f52ff8d4ab2b9faa30862f82565aec244d854a2b9a1a3e25307f891bdc"
    },
    {
      "path": "results/hand_labels/batch2/batch2.csv",
      "sha256": "e436534af8e0358ae31348a9e3ab3606118ef55f282ba0b901838d4bae33a3a0"
    },
    {
      "path": "results/division_p1/deploy_v4_modern_alltrain/DIVISION_VERIFIER_CONFIG.json",
      "sha256": "e0107bb1743cc814bcb9916e1a679585c3016e49ca0441bf6bccc34c9fe92c06"
    },
    {
      "path": "results/division_p1/deploy_v5_hand_p1/DIVISION_VERIFIER_CONFIG.json",
      "sha256": "8064e2732a0921aaed4338585e194db2a4a379d207ed2580a372c02082d9ddcb"
    },
    {
      "path": "results/division_p1/deploy_v6_hand_b2/DIVISION_VERIFIER_CONFIG.json",
      "sha256": "494d2989c788141866ec90cac2e6b664a341b064b0584191fe963a41358a4e0a"
    },
    {
      "path": "results/division_p1/deploy_v7_v2feat_b2/DIVISION_VERIFIER_CONFIG.json",
      "sha256": "6940ac7d47ab115bb8a3c88bda03ac25b95f222d85bad9a87fcc52e5a5a5eb66"
    },
    {
      "path": "results/division_p1/deploy_v5_hand_partial250/INGEST_REPORT.json",
      "sha256": "af8b3a340b31fa7422dd1e51e49aadf98c84404e16e383c8f0f253ceda5d0b1d"
    },
    {
      "path": "results/division_p1/deploy_v5_hand_batch2only/INGEST_REPORT.json",
      "sha256": "0014089a3eb2bdd5472598655865eb0ab84177df52602ecc3af9e1b01fe69851"
    },
    {
      "path": "results/division_p1/v85_oof_evidence.json",
      "sha256": "b9a11d347502e82850a9f57c86d9d0cc11b22153a7d2ea0586bbe7dcf8a729da"
    },
    {
      "path": "results/division_p1/v87_oof_evidence.json",
      "sha256": "a78b0ad5b7507ccf7af92bae7e24273e351a0a8d1cb0b7081451fa3a3ea2d69c"
    },
    {
      "path": "results/division_p1/missed_gt_diag_v1/report.json",
      "sha256": "4477eab1b588d00c750487e03889d081dad6cb195a6d33c5e4fbe583fc5ba08a"
    },
    {
      "path": "results/hand_labels/batch1/hand_labels_batch1_final.json",
      "sha256": "ca4cbf1799171cf76829df407d1f118c84b2d230e4f0112cce790094de862ec0"
    },
    {
      "path": "results/hand_labels/batch2/hand_labels_batch2_final.json",
      "sha256": "3d5211acc40fddc31e9851da22bbeb53b6d21afa8d1adfbdf681891cd79514fd"
    },
    {
      "path": "results/swap_adjudication_20260914/preparation/CARD_S4.json",
      "sha256": "2a5f05d6e2ce26c5b8a5407ab8773d93b19e198f1d3e72f3841c071d64a311ac"
    },
    {
      "path": "results/swap_adjudication_20260914/batch1/RESULT.json",
      "sha256": "ba238833a4d1688ab300f3c8afa9e12809bb9e9f68280409356f141f4091327b"
    },
    {
      "path": "results/swap_adjudication_20260914/batch1/labels/LABELER_REPORT.json",
      "sha256": "76089c837a30bc3cd7c7653be9788ab5870a272171660e89ccdeeb6232bc3f9f"
    },
    {
      "path": "results/swap_adjudication_20260914/batch1/EXPLORATORY_BETWEEN_GEOMETRY.json",
      "sha256": "b320d5b5d906193b6460ab631be35cc1d8b9f9b8e24428be089e7ec15f734fe0"
    },
    {
      "path": "docs/CLAUDE_SESSION_LOG.md",
      "sha256": "0d774b674ec0547bcdd34c4ffb3d142c24eda6d2ac3a7f93d194dda6907c9294"
    },
    {
      "path": "results/kaggle_outputs/v87/division_verifier_v7_runtime/runtime/biohub/division_verifier_runtime.py",
      "sha256": "d16d5686dff739f2688f7d5f16a3e08c7bf8b72e39a6623a2323aa13e3e65ede"
    }
  ]
}