# Independent Em0034 Review Receipt

- Object ID: `em:research-note:sha256:7526f1929330c0f2ecd06d2f20c0864c239235c075d08ed99920c858980d91c7`
- Kind: `research-note`
- Repository path: [`research/how-we-know/gpt-4-bar-exam-percentile/independent-em0034-review-receipt.json`](https://github.com/yoheinakajima/epistemedia/blob/f92846570180dfa4511263f8ba98ecd18f7772c9/research/how-we-know/gpt-4-bar-exam-percentile/independent-em0034-review-receipt.json)
- Content digest: `f35f8c093778ad1cdeafc57746575f1a82d12fe6e6fbf0d1d4a546e3c5296e1e`

**Also filed under:** [Research Program](https://epistemedia.org/topics/research-program/)

## Source content

{
  "format": "epistemedia-independent-dossier-review-v1",
  "task_id": "EM-0034",
  "repository": "https://github.com/yoheinakajima/epistemedia",
  "reviewer": {
    "id": "codex-independent-em0034-reviewer",
    "independent": true,
    "authored_candidate": false,
    "fresh_clone": true,
    "model_family": "OpenAI GPT-5"
  },
  "decision": "pass",
  "complete": true,
  "started_at": "2026-08-28T15:57:21Z",
  "completed_at": "2026-08-28T15:59:23Z",
  "reviewed": {
    "base": "ead2b5887da233dde4e790f0c6c0aea15e59f074",
    "head": "1c5a8318b2bb770c38a07a96f731531ba1c53a32",
    "tree": "b7356ab1578e8a1eea76cb14e36d7a15b0e00d48",
    "dossier": {
      "bytes": 224792,
      "path": "research/how-we-know/gpt-4-bar-exam-percentile/candidate-dossier.json",
      "sha256": "32c4457b3823237b2f988a26d51b2f6222af8060e662993524aff1c1a5d79e5d",
      "dossier_id": "em:dossier:sha256:babe89ba3bda594a8d9f2db86a5a2987f284437a069b940d19b6928856d936d1"
    },
    "source_graph_sha256": "2bb896ded74d63469ec4e3947d75adfd6ac9a9c5dee6fe0db7ef9251760da38e"
  },
  "accepted_bindings": [
    {
      "bytes": 210543,
      "path": "research/how-we-know/gpt-4-bar-exam-percentile/candidate-packet.json",
      "sha256": "07dec44bc2cf893b19ba6307e34aef60b6127b5e881e319ff4132c8e69972d1a"
    },
    {
      "bytes": 92747,
      "path": "research/how-we-know/gpt-4-bar-exam-percentile/independent-review-receipt.json",
      "sha256": "0606e98ed7ff98f1d0e69509db5edcd8dce774c62c48981bd5aec8f23a95ad20"
    },
    {
      "bytes": 64382,
      "path": "research/how-we-know/gpt-4-bar-exam-percentile/source-records.json",
      "sha256": "385bbc193be495597ef0fe2a675220ca443297e673b32947c8d8ffcd1a6366da"
    },
    {
      "bytes": 62565,
      "path": "research/how-we-know/gpt-4-bar-exam-percentile/artifact-inventory.json",
      "sha256": "f9e45045831a15fcb40c7f542c73651e5dc63ba081b145d00260bee91192c25b"
    },
    {
      "bytes": 51950,
      "path": "research/how-we-know/gpt-4-bar-exam-percentile/git-blob-search-manifest.json",
      "sha256": "708657f61bc759f3a5b439affde1a677177859b50a8bc960012b81dcebcdb122"
    }
  ],
  "coverage": {
    "assertion_keys": [
      "assertion-claim-february-sensitive",
      "assertion-claim-july-sensitive",
      "assertion-claim-launch-comparison-unspecified",
      "assertion-claim-launch-score-label",
      "assertion-claim-martinez-first-time",
      "assertion-claim-martinez-passers-conflict",
      "assertion-claim-no-lawyer-rank",
      "assertion-claim-score-discrepancy",
      "assertion-derive-illinois-feb-2018-298",
      "assertion-derive-illinois-feb-2019-298",
      "assertion-derive-illinois-jul-2018-298",
      "assertion-derive-martinez-first-time-essay",
      "assertion-derive-martinez-first-time-mbe",
      "assertion-derive-martinez-first-time-ube",
      "assertion-derive-martinez-parameters",
      "assertion-derive-martinez-passers-essay",
      "assertion-derive-martinez-passers-mbe",
      "assertion-derive-martinez-passers-ube",
      "assertion-encyclopedia-evaluation",
      "assertion-reviewed-source-register",
      "assertion-skeptical-evaluation"
    ],
    "calculation_ids": [
      "derive-illinois-feb-2018-298",
      "derive-illinois-feb-2019-298",
      "derive-illinois-jul-2018-298",
      "derive-martinez-first-time-essay",
      "derive-martinez-first-time-mbe",
      "derive-martinez-first-time-ube",
      "derive-martinez-parameters",
      "derive-martinez-passers-essay",
      "derive-martinez-passers-mbe",
      "derive-martinez-passers-ube"
    ],
    "edition_keys": [
      "edition-em0032-calculation-register",
      "edition-illinois-feb-2018",
      "edition-illinois-feb-2019",
      "edition-illinois-jul-2018",
      "edition-katz-figshare-25018513-v1",
      "edition-katz-git-commit-90997f7-tree-810bd4a",
      "edition-katz-rsta-2024",
      "edition-katz-ssrn-4389233",
      "edition-martinez-2024-vor",
      "edition-martinez-osf-analysis-new",
      "edition-martinez-osf-c8ygu",
      "edition-ncbe-first-repeat-2022",
      "edition-ncbe-mbe-2022",
      "edition-ncbe-snapshot-2022",
      "edition-ncbe-ube-2022",
      "edition-ncbe-ube-scores-2026-08-27",
      "edition-ny-passrates-2022",
      "edition-openai-2303-08774v1",
      "edition-openai-2303-08774v6",
      "edition-reshetar-testing-column-spring-2022"
    ],
    "lineage_edge_ids": [
      "edge-author-social-collaboration",
      "edge-benchmark-illinois-charts",
      "edge-citation-katz-to-martinez",
      "edge-comparison-class-first-time-passers",
      "edge-data-reported-score-reuse",
      "edge-derivation-comparison-inputs",
      "edge-material-shared-exam-items",
      "edge-method-single-free-response-run",
      "edge-model-historical-snapshots",
      "edge-score-component-composite"
    ],
    "lineage_keys": [
      "lineage-evaluation-synthesis",
      "lineage-illinois-comparison-root",
      "lineage-martinez-analysis-root",
      "lineage-model-performance-root",
      "lineage-ncbe-comparison-root",
      "lineage-new-york-pass-rate-root",
      "lineage-reviewed-source-register"
    ],
    "proposition_keys": [
      "claim-february-sensitive",
      "claim-july-sensitive",
      "claim-launch-comparison-unspecified",
      "claim-launch-score-label",
      "claim-martinez-first-time",
      "claim-martinez-passers-conflict",
      "claim-no-lawyer-rank",
      "claim-score-discrepancy",
      "derive-illinois-feb-2018-298",
      "derive-illinois-feb-2019-298",
      "derive-illinois-jul-2018-298",
      "derive-martinez-first-time-essay",
      "derive-martinez-first-time-mbe",
      "derive-martinez-first-time-ube",
      "derive-martinez-parameters",
      "derive-martinez-passers-essay",
      "derive-martinez-passers-mbe",
      "derive-martinez-passers-ube",
      "prop-encyclopedia-evaluation",
      "prop-reviewed-source-register",
      "prop-skeptical-evaluation"
    ],
    "relation_keys": [
      "edge-author-social-collaboration",
      "edge-benchmark-illinois-charts",
      "edge-citation-katz-to-martinez",
      "edge-comparison-class-first-time-passers--1",
      "edge-comparison-class-first-time-passers--2",
      "edge-data-reported-score-reuse",
      "edge-derivation-comparison-inputs--1",
      "edge-derivation-comparison-inputs--2",
      "edge-material-shared-exam-items",
      "edge-method-single-free-response-run",
      "edge-model-historical-snapshots",
      "edge-score-component-composite",
      "relation-assertion-claim-february-sensitive",
      "relation-assertion-claim-july-sensitive",
      "relation-assertion-claim-launch-comparison-unspecified",
      "relation-assertion-claim-launch-score-label",
      "relation-assertion-claim-martinez-first-time",
      "relation-assertion-claim-martinez-passers-conflict",
      "relation-assertion-claim-no-lawyer-rank",
      "relation-assertion-claim-score-discrepancy",
      "relation-assertion-derive-illinois-feb-2018-298",
      "relation-assertion-derive-illinois-feb-2019-298",
      "relation-assertion-derive-illinois-jul-2018-298",
      "relation-assertion-derive-martinez-first-time-essay",
      "relation-assertion-derive-martinez-first-time-mbe",
      "relation-assertion-derive-martinez-first-time-ube",
      "relation-assertion-derive-martinez-parameters",
      "relation-assertion-derive-martinez-passers-essay",
      "relation-assertion-derive-martinez-passers-mbe",
      "relation-assertion-derive-martinez-passers-ube",
      "relation-assertion-encyclopedia-evaluation",
      "relation-assertion-reviewed-source-register",
      "relation-assertion-skeptical-evaluation"
    ],
    "source_work_keys": [
      "work-em0032-calculation-register",
      "work-illinois-percentile-charts",
      "work-katz-gpt4-bar",
      "work-martinez-reanalysis",
      "work-ncbe-2022-statistics",
      "work-ncbe-ube",
      "work-ny-2022-passrates",
      "work-openai-gpt4-report",
      "work-reshetar-testing-column-february-july"
    ],
    "span_keys": [
      "span-calculation-derive-illinois-feb-2018-298",
      "span-calculation-derive-illinois-feb-2019-298",
      "span-calculation-derive-illinois-jul-2018-298",
      "span-calculation-derive-martinez-first-time-essay",
      "span-calculation-derive-martinez-first-time-mbe",
      "span-calculation-derive-martinez-first-time-ube",
      "span-calculation-derive-martinez-parameters",
      "span-calculation-derive-martinez-passers-essay",
      "span-calculation-derive-martinez-passers-mbe",
      "span-calculation-derive-martinez-passers-ube",
      "span-illinois-feb-2018-anchors",
      "span-illinois-feb-2019-anchors",
      "span-illinois-jul-2018-anchors",
      "span-katz-vor-abstract-score",
      "span-katz-vor-components",
      "span-katz-vor-materials",
      "span-katz-vor-percentile-boundary",
      "span-katz-vor-range",
      "span-katz-vor-score-discrepancy",
      "span-martinez-abstract-ranks",
      "span-martinez-discussion-48",
      "span-martinez-mean-assumption",
      "span-martinez-model-assumptions",
      "span-martinez-results-45",
      "span-martinez-score-validation",
      "span-martinez-script-comment-conflict",
      "span-martinez-script-july-mbe-distribution",
      "span-martinez-script-thresholds",
      "span-martinez-script-ube-sd",
      "span-martinez-table-45",
      "span-ncbe-jurisdiction-status",
      "span-ncbe-mbe-2022-counts",
      "span-ncbe-ny-cutoff-2022",
      "span-ncbe-snapshot-composition",
      "span-ncbe-snapshot-typo",
      "span-ncbe-ube-weights",
      "span-ny-first-timers-2022",
      "span-openai-v1-abstract-top-ten",
      "span-openai-v1-collaborators",
      "span-openai-v1-free-response-run",
      "span-openai-v1-scoring",
      "span-openai-v1-snapshots",
      "span-openai-v1-table-score",
      "span-openai-v6-table-score",
      "span-reshetar-first-time-mean"
    ]
  },
  "commands": [
    {
      "argv": [
        "python",
        "research/how-we-know/gpt-4-bar-exam-percentile/build_candidate.py",
        "--check"
      ],
      "exit_code": 0,
      "stdout_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
    },
    {
      "argv": [
        "python",
        "research/how-we-know/gpt-4-bar-exam-percentile/verify_candidate.py",
        "--self-test"
      ],
      "exit_code": 0,
      "stdout_sha256": "42e3a9b3c109f75a761a1a829a904af0224da51d69b51221c68a9cb30108becc"
    },
    {
      "argv": [
        "python",
        "-m",
        "pytest",
        "-q",
        "research/how-we-know/gpt-4-bar-exam-percentile/test_candidate.py"
      ],
      "exit_code": 0,
      "stdout_sha256": "585ba42829657b9e2d56f8614be6c3365a8405474eed008559511ab50f2ffcfb"
    },
    {
      "argv": [
        "ruff",
        "check",
        "research/how-we-know/gpt-4-bar-exam-percentile/build_candidate.py",
        "research/how-we-know/gpt-4-bar-exam-percentile/verify_candidate.py",
        "research/how-we-know/gpt-4-bar-exam-percentile/test_candidate.py"
      ],
      "exit_code": 0,
      "stdout_sha256": "82b3e6a6c090a57601d22943bd23fca9218d1031dbe5a7b754092f9a156b4f18"
    },
    {
      "argv": [
        "git",
        "diff",
        "--check",
        "ead2b5887da233dde4e790f0c6c0aea15e59f074..1c5a8318b2bb770c38a07a96f731531ba1c53a32"
      ],
      "exit_code": 0,
      "stdout_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
    },
    {
      "argv": [
        "make",
        "check",
        "PYTHON=/private/tmp/em0034-independent-review.v8smBm/repo/.venv/bin/python"
      ],
      "exit_code": 0,
      "stdout_sha256": "9e45d714ee36459e11298f1864b94534c65fa5eacea8f3d74bba13362638be4a"
    }
  ],
  "limitations": [
    "This review verifies the bounded historical candidate and does not establish current model behavior, general legal competence, practicing-lawyer quality, or a current product ranking.",
    "The launch comparison distribution, original top-ten-percent chart, national UBE distribution, and proprietary inputs remain unresolved and receive no invented value or evidentiary credit.",
    "The Martinez re-analysis remains model- and assumption-dependent, including its retained 45/48 internal discrepancy and its separate comparison populations.",
    "This receipt does not admit, feature, publish, deploy, or describe Case 003 as live."
  ],
  "findings": [
    "Accepted EM-0032 artifacts and receipts are byte-identical to accepted commit 700a822f38d00d13cc0661fd577bdb7e6e5b34dd.",
    "All ten accepted derivations close through exact structured calculation spans, parent spans, and resolved input-cell identities.",
    "All ten accepted typed dependence-edge groups retain exact evidence and every mapped structural endpoint.",
    "The 297/298 and 45/48 discrepancies, comparison-class boundaries, historical-model boundary, and lawyer-comparator limitation remain visible.",
    "Encyclopedia and skeptical evaluations use one unchanged source graph with materially different reasons and reader upshots.",
    "Exact-head CI run 33187448187 and job 98904022930 completed successfully with no review threads."
  ],
  "git_state": {
    "fresh_clone": true,
    "pre_review_clean": true,
    "post_review_clean": true,
    "unchanged_during_review": true,
    "pre_review_head": "1c5a8318b2bb770c38a07a96f731531ba1c53a32",
    "post_review_head": "1c5a8318b2bb770c38a07a96f731531ba1c53a32",
    "receipt_path": "research/how-we-know/gpt-4-bar-exam-percentile/independent-em0034-review-receipt.json"
  }
}