Publish WorkLedger Structured Edit Interface Study v0.1.0

DATASET_CARD.md

@@ -0,0 +1,53 @@
+# WorkLedger Structured Edit Interface Study v0.1
+
+Author: Confetti3
+
+This upload root is the offline release-candidate dataset card for `workledger-structured-edit-interface-study-v0.1`, destined for `confetti3/workledger-structured-edit-interface-study-v0.1`. The Hugging Face repository is hosted under the authenticated namespace `confetti3`, while public authorship is credited only to the pseudonym Confetti3. It is built from the sanitized public package under `publications/workledger-structured-edit-interface-study-v0.1`; it is not a live runner, it constructs no provider transport, and it performs zero network or credential operations.
+
+## Study scope
+
+WorkLedger v0.1 evaluates structured editing interfaces for language-model-generated code edits under fail-closed validation. The package contains sanitized final assistant answers, task metadata, aggregate outcomes, claim-survival records, schemas, provenance, leakage checks, and deterministic manifests. It supports public inspection of the controlled interface study records; it does not include hidden evaluator source, raw provider bodies, request identifiers, reasoning traces, credentials, SQLite ledgers, or private artifact paths.
+
+## Registered findings
+
+- Exact replacement advantage replicated for Qwen3 14B: U1 1/19 and R 13/19 on the registered paired primary comparison.
+- Exact replacement advantage replicated for Qwen3 235B A22B: U1 3/19 and R 16/19 on the registered paired primary comparison.
+- Qwen3-32B is an external anchor from existing calibration evidence: U1 0/19 and R 9/19 in the same direction, not jointly randomized and not in the paid multimodel denominator.
+- Qwen3 8B was directionally favorable to R, with U1 2/19 and R 5/19, but paired coverage was 18/19 and the registered interface-advantage rule was not met.
+- DeepSeek V4 Pro is infrastructure-unassessable: 38/38 cells were provider-route privacy/data-policy HTTP 404 failures with empty final answers. This is not a coding-performance score and must not be treated as a model-performance denominator.
+- `CROSS_MODEL_GENERALIZATION_SIGNAL` is false. The model-strength interaction is descriptive only, not a parameter-scaling law.
+- Strict native and one-outer-fence-normalized policies are reported on identical response bytes and were identical in the new multimodel run.
+
+## Limitations
+
+The evidence uses exactly one attempt per cell. Provider routes and inference parameters are confounded with model identity. The task bank is calibrated and bounded; it is not a production-readiness claim and should not be generalized to broad software engineering workloads. DeepSeek V4 Pro produced no valid model outputs in this run, so it supports neither the interface advantage nor its refutation.
+
+## Ethics, privacy, and security
+
+The public package is sanitized by field selection and leakage scans. Private execution evidence is withheld, including authorization material, request envelopes, response identifiers, raw provider bodies, reasoning fields, hidden evaluator source, private filesystem paths, and SQLite ledgers. The release candidate includes text-only upload rehearsal files, but no upload, tag, release, DOI, repository creation, provider request, or credential access has been executed.
+
+## Data structure
+
+- `data/generations.jsonl`: sanitized calibration and multimodel final assistant answers plus public outcome metadata.
+- `data/tasks.jsonl`: public task metadata and hashes with source/evaluator bytes withheld.
+- `data/aggregate-results.json` and `data/aggregate_results.jsonl`: immutable aggregate outcomes.
+- `data/claim-survival.json` and `data/claim_survival.jsonl`: audited public claims and safest wording.
+- `schema/*.schema.json`: public record schemas.
+- `provenance/*.json`: build provenance, manifest aliases, leakage checks, sanitization proof, license audit, and withheld-content policy.
+- `MANIFEST.json` and `MANIFEST.sha256`: release-root integrity manifest and digest.
+
+## Provenance and task wording
+
+Low-affinity calibration tasks are described as newly authored for calibration when supported by release provenance. High-affinity tasks are described as adapted or reconstructed from upstream issue/task patterns; upstream source bytes and evaluator bytes are withheld. The package preserves frozen hashes, pass counts, and decision labels.
+
+## Citation
+
+Please cite as: Confetti3. WorkLedger Structured Edit Interface Study v0.1. Version 0.1.0. Release date 2026-07-20.
+
+## License treatment
+
+Public dataset metadata and public release records are licensed under CC-BY-4.0. Space code and tooling are licensed under MIT. This release does not relicense upstream source bytes, evaluator bytes, private execution evidence, or withheld fixtures.
+
+## Withheld-content policy
+
+The release discloses sanitized records and non-sensitive hashes only. Hidden evaluators, raw provider evidence, request/response identifiers, reasoning traces, credentials, private paths, SQLite ledgers, and upstream source/evaluator bytes are withheld.

MANIFEST.json

@@ -0,0 +1,169 @@
+{
+  "entries": [
+    {
+      "bytes": 15058,
+      "path": "data/aggregate_results.jsonl",
+      "sha256": "81bf95b543e8d474c14c7b350978fce7bf541f0940c7720ee2ab89c40ff63c87"
+    },
+    {
+      "bytes": 20118,
+      "path": "data/aggregate-results.json",
+      "sha256": "6c8daea0e22c7aad2a93a37a533ebc571193ab01301e4eda329293469d57ca38"
+    },
+    {
+      "bytes": 10346,
+      "path": "data/claim_survival.jsonl",
+      "sha256": "3341108d1e51f0b1a249e7e7fdc9edce7fef59598e4817b4a1364964236a49c9"
+    },
+    {
+      "bytes": 11765,
+      "path": "data/claim-survival.json",
+      "sha256": "45aa24c6a3c0676fe888e6aa49893f4c05b42455e115342aa64cc3b738fa0c94"
+    },
+    {
+      "bytes": 667543,
+      "path": "data/generations.jsonl",
+      "sha256": "78363d7c143d289b092dde2c1ae1c674529d621573ff5c53d4c5618984b00299"
+    },
+    {
+      "bytes": 3177,
+      "path": "data/model-metadata.json",
+      "sha256": "16acadc395b0f211b65bae3ad89ea94675ec6645b16e73a6d6de11ad1bd8b628"
+    },
+    {
+      "bytes": 667543,
+      "path": "data/stage-outcomes.jsonl",
+      "sha256": "78363d7c143d289b092dde2c1ae1c674529d621573ff5c53d4c5618984b00299"
+    },
+    {
+      "bytes": 1165,
+      "path": "data/studies.jsonl",
+      "sha256": "80a2c92bc7657c91405f1fc6d89a33fe2b5e4547b22539fa83baadee90cbdc6d"
+    },
+    {
+      "bytes": 1165,
+      "path": "data/study_metadata.jsonl",
+      "sha256": "80a2c92bc7657c91405f1fc6d89a33fe2b5e4547b22539fa83baadee90cbdc6d"
+    },
+    {
+      "bytes": 46642,
+      "path": "data/tasks.jsonl",
+      "sha256": "416b1602fd0aedd2c34274f7752b5c7864b5f8a4ea07c47de9c666e9561390ca"
+    },
+    {
+      "bytes": 5029,
+      "path": "DATASET_CARD.md",
+      "sha256": "f6467f87d591bb46927195c40019eedd516352baacd10d48931e5760b441354c"
+    },
+    {
+      "bytes": 896,
+      "path": "docs/limitations.md",
+      "sha256": "08644dd658ffa1fb0a99984ca869516eb47eacf595096eea1f939d67c7a409fb"
+    },
+    {
+      "bytes": 250,
+      "path": "docs/LIMITATIONS.md",
+      "sha256": "d9ccebf4871254c2b9ad8373adbdc4edf9d06ef848c4dd1dde5b375bee3237d4"
+    },
+    {
+      "bytes": 281,
+      "path": "docs/METHODOLOGY.md",
+      "sha256": "51b4a0c88206d806668e75200c4e248eb17b8788e9481931210e688b0709f886"
+    },
+    {
+      "bytes": 743,
+      "path": "docs/methods.md",
+      "sha256": "5a162830489255a4bf9f8dfbc59f5dc66abd83f84a531a65fe5e4eda5c1d4c51"
+    },
+    {
+      "bytes": 283,
+      "path": "docs/REPRODUCTION.md",
+      "sha256": "1ba152f73c84ebb3ade2455ec38fff7b1c13709bb1c6e825569de3e6ff8d35c2"
+    },
+    {
+      "bytes": 306,
+      "path": "docs/SECURITY_AND_PRIVACY.md",
+      "sha256": "169a3738a8e4f4ceaf4388012b02111ffdefd6e7c915608caaa1949ff8b6912f"
+    },
+    {
+      "bytes": 170,
+      "path": "docs/suggested-tag.md",
+      "sha256": "519e311ce283705d6dcbbbb3ed89be15a1d94ebfd83f646f28a188b5da33208f"
+    },
+    {
+      "bytes": 997,
+      "path": "provenance/BUILD_PROVENANCE.json",
+      "sha256": "c1711b42e31a56809962cb2499009af23575434dc3f3fd4a272297f9aae47fbb"
+    },
+    {
+      "bytes": 14138,
+      "path": "provenance/LICENSE_AUDIT.json",
+      "sha256": "0f78166ae466a075367b4b2e08eb1f097dd784a27c5ddf680dc75d737536686c"
+    },
+    {
+      "bytes": 5222,
+      "path": "provenance/MANIFEST.json",
+      "sha256": "aac6136a284b3b4af786b145e074de48e2ead35bc38390cc556acdd9fe031180"
+    },
+    {
+      "bytes": 997,
+      "path": "provenance/PROVENANCE.json",
+      "sha256": "c1711b42e31a56809962cb2499009af23575434dc3f3fd4a272297f9aae47fbb"
+    },
+    {
+      "bytes": 394,
+      "path": "provenance/PUBLIC_EXPORT_LEAKAGE_CHECK.json",
+      "sha256": "e262a6495c3a7548822367e88d12268148d6458a84dc66a9b63d553c9f538d52"
+    },
+    {
+      "bytes": 4705,
+      "path": "provenance/PUBLIC_EXPORT_MANIFEST.json",
+      "sha256": "a2ec95e3493df2284c6fed1d4a7e2c510729794e64914095edba45731a30a4e4"
+    },
+    {
+      "bytes": 65,
+      "path": "provenance/PUBLIC_EXPORT_MANIFEST.sha256",
+      "sha256": "84a8f3f46de699c1c618c6d8fe46cf477b6ad0ad99636f782f1847afa16328b4"
+    },
+    {
+      "bytes": 1082,
+      "path": "provenance/PUBLIC_EXPORT_SANITIZATION_PROOF.json",
+      "sha256": "525960c306a747a64e425698676c392097135cbd7deb0744054070ad8b04af33"
+    },
+    {
+      "bytes": 704,
+      "path": "provenance/WITHHELD_COMPONENT_HASHES.json",
+      "sha256": "3f25c6a4b22cdcf8e3be858c6657cc238c3e1236aafd7d05eb68902518298230"
+    },
+    {
+      "bytes": 5649,
+      "path": "README.md",
+      "sha256": "2a729ae7215e2bfb59eafeed9003d5de499bad19be808c00440a06c85fcd65fb"
+    },
+    {
+      "bytes": 765,
+      "path": "RELEASE_NOTES.md",
+      "sha256": "083d6955e7d4a4274baa322f382312cb047bdc8e4e411d3976f1fdf8ea8fbe32"
+    },
+    {
+      "bytes": 1217,
+      "path": "schema/generation-record.schema.json",
+      "sha256": "c5c3159dafd793ff9e646a11c4c1f23c8076eff17889a9acd4349c80700375c4"
+    },
+    {
+      "bytes": 441,
+      "path": "schema/public-export.schema.json",
+      "sha256": "dc8211ef99efe097b2de9efa593cdb867bed94b85963e027328d0ce26c9a2998"
+    },
+    {
+      "bytes": 615,
+      "path": "schema/task-record.schema.json",
+      "sha256": "68272ec4c35504fdeadd766bd0ce4fed44b9ea5cfedd2a895a4e14b3121d742c"
+    }
+  ],
+  "generatedOffline": true,
+  "packageRoot": "release/huggingface-dataset",
+  "providerRequests": 0,
+  "recordKind": "WORKLEDGER_V0_1_HF_DATASET_MANIFEST",
+  "schemaVersion": 1
+}

MANIFEST.sha256

@@ -0,0 +1 @@
+1c88282f96377f52e643686fe5b92402e9cff9562a4d61d3f938c1fedc380878

README.md

@@ -0,0 +1,81 @@
+---
+pretty_name: WorkLedger Structured Edit Interface Study v0.1
+license: cc-by-4.0
+tags:
+  - structured-editing
+  - structured-edit-interface-study
+  - language-model-interfaces
+  - verified-code-edits
+  - fail-closed-evaluation
+  - software-engineering-agents
+  - controlled-interface-study
+task_categories:
+  - text-generation
+language:
+  - en
+size_categories:
+  - n<1K
+configs:
+  - config_name: default
+    data_files:
+      - split: generations
+        path: data/generations.jsonl
+      - split: tasks
+        path: data/tasks.jsonl
+      - split: aggregate_results
+        path: data/aggregate_results.jsonl
+---
+
+# WorkLedger Structured Edit Interface Study v0.1
+
+Author: Confetti3
+
+This upload root is the offline release-candidate dataset card for `workledger-structured-edit-interface-study-v0.1`, destined for `confetti3/workledger-structured-edit-interface-study-v0.1`. The Hugging Face repository is hosted under the authenticated namespace `confetti3`, while public authorship is credited only to the pseudonym Confetti3. It is built from the sanitized public package under `publications/workledger-structured-edit-interface-study-v0.1`; it is not a live runner, it constructs no provider transport, and it performs zero network or credential operations.
+
+## Study scope
+
+WorkLedger v0.1 evaluates structured editing interfaces for language-model-generated code edits under fail-closed validation. The package contains sanitized final assistant answers, task metadata, aggregate outcomes, claim-survival records, schemas, provenance, leakage checks, and deterministic manifests. It supports public inspection of the controlled interface study records; it does not include hidden evaluator source, raw provider bodies, request identifiers, reasoning traces, credentials, SQLite ledgers, or private artifact paths.
+
+## Registered findings
+
+- Exact replacement advantage replicated for Qwen3 14B: U1 1/19 and R 13/19 on the registered paired primary comparison.
+- Exact replacement advantage replicated for Qwen3 235B A22B: U1 3/19 and R 16/19 on the registered paired primary comparison.
+- Qwen3-32B is an external anchor from existing calibration evidence: U1 0/19 and R 9/19 in the same direction, not jointly randomized and not in the paid multimodel denominator.
+- Qwen3 8B was directionally favorable to R, with U1 2/19 and R 5/19, but paired coverage was 18/19 and the registered interface-advantage rule was not met.
+- DeepSeek V4 Pro is infrastructure-unassessable: 38/38 cells were provider-route privacy/data-policy HTTP 404 failures with empty final answers. This is not a coding-performance score and must not be treated as a model-performance denominator.
+- `CROSS_MODEL_GENERALIZATION_SIGNAL` is false. The model-strength interaction is descriptive only, not a parameter-scaling law.
+- Strict native and one-outer-fence-normalized policies are reported on identical response bytes and were identical in the new multimodel run.
+
+## Limitations
+
+The evidence uses exactly one attempt per cell. Provider routes and inference parameters are confounded with model identity. The task bank is calibrated and bounded; it is not a production-readiness claim and should not be generalized to broad software engineering workloads. DeepSeek V4 Pro produced no valid model outputs in this run, so it supports neither the interface advantage nor its refutation.
+
+## Ethics, privacy, and security
+
+The public package is sanitized by field selection and leakage scans. Private execution evidence is withheld, including authorization material, request envelopes, response identifiers, raw provider bodies, reasoning fields, hidden evaluator source, private filesystem paths, and SQLite ledgers. The release candidate includes text-only upload rehearsal files, but no upload, tag, release, DOI, repository creation, provider request, or credential access has been executed.
+
+## Data structure
+
+- `data/generations.jsonl`: sanitized calibration and multimodel final assistant answers plus public outcome metadata.
+- `data/tasks.jsonl`: public task metadata and hashes with source/evaluator bytes withheld.
+- `data/aggregate-results.json` and `data/aggregate_results.jsonl`: immutable aggregate outcomes.
+- `data/claim-survival.json` and `data/claim_survival.jsonl`: audited public claims and safest wording.
+- `schema/*.schema.json`: public record schemas.
+- `provenance/*.json`: build provenance, manifest aliases, leakage checks, sanitization proof, license audit, and withheld-content policy.
+- `MANIFEST.json` and `MANIFEST.sha256`: release-root integrity manifest and digest.
+
+## Provenance and task wording
+
+Low-affinity calibration tasks are described as newly authored for calibration when supported by release provenance. High-affinity tasks are described as adapted or reconstructed from upstream issue/task patterns; upstream source bytes and evaluator bytes are withheld. The package preserves frozen hashes, pass counts, and decision labels.
+
+## Citation
+
+Please cite as: Confetti3. WorkLedger Structured Edit Interface Study v0.1. Version 0.1.0. Release date 2026-07-20.
+
+## License treatment
+
+Public dataset metadata and public release records are licensed under CC-BY-4.0. Space code and tooling are licensed under MIT. This release does not relicense upstream source bytes, evaluator bytes, private execution evidence, or withheld fixtures.
+
+## Withheld-content policy
+
+The release discloses sanitized records and non-sensitive hashes only. Hidden evaluators, raw provider evidence, request/response identifiers, reasoning traces, credentials, private paths, SQLite ledgers, and upstream source/evaluator bytes are withheld.

RELEASE_NOTES.md

@@ -0,0 +1,7 @@
+# Release notes: WorkLedger Structured Edit Interface Study v0.1
+
+Status: READY_AWAITING_PUBLIC_UPLOAD_AUTHORIZATION.
+
+This release candidate preserves the frozen public results while correcting citation, Zenodo, provenance, licensing, and terminology wording for an offline public package authored as Confetti3. No tag, GitHub release, Hugging Face upload, Zenodo upload, DOI, repository creation, provider request, or credential access has been performed.
+
+Key preserved results: Qwen3 14B U1 1/19 R 13/19; Qwen3 235B A22B U1 3/19 R 16/19; Qwen3-32B external anchor U1 0/19 R 9/19; Qwen3 8B U1 2/19 R 5/19 with paired coverage 18/19 and no registered interface advantage; DeepSeek V4 Pro infrastructure-unassessable with 38/38 privacy/data-policy route failures.

data/aggregate-results.json

@@ -0,0 +1,623 @@
+[
+  {
+    "arms": {
+      "R": {
+        "applicable": 9,
+        "fenceRemoved": 0,
+        "hiddenPasses": 9,
+        "lfAppended": 0,
+        "n": 19,
+        "validEnvelopes": 11,
+        "visiblePasses": 9,
+        "wilsonHidden": {
+          "lower95": 0.2732980516117956,
+          "method": "wilson_score",
+          "n": 19,
+          "proportion": 0.47368421052631576,
+          "successes": 9,
+          "upper95": 0.6829219075488682
+        }
+      },
+      "RN": {
+        "applicable": 15,
+        "fenceRemoved": 11,
+        "hiddenPasses": 14,
+        "lfAppended": 0,
+        "n": 19,
+        "validEnvelopes": 18,
+        "visiblePasses": 14,
+        "wilsonHidden": {
+          "lower95": 0.5120844910218489,
+          "method": "wilson_score",
+          "n": 19,
+          "proportion": 0.7368421052631579,
+          "successes": 14,
+          "upper95": 0.8819358765321749
+        }
+      },
+      "U0": {
+        "applicable": 0,
+        "fenceRemoved": 0,
+        "hiddenPasses": 0,
+        "lfAppended": 0,
+        "n": 19,
+        "validEnvelopes": 0,
+        "visiblePasses": 0,
+        "wilsonHidden": {
+          "lower95": 0,
+          "method": "wilson_score",
+          "n": 19,
+          "proportion": 0,
+          "successes": 0,
+          "upper95": 0.16817922405261626
+        }
+      },
+      "U1": {
+        "applicable": 0,
+        "fenceRemoved": 0,
+        "hiddenPasses": 0,
+        "lfAppended": 1,
+        "n": 19,
+        "validEnvelopes": 0,
+        "visiblePasses": 0,
+        "wilsonHidden": {
+          "lower95": 0,
+          "method": "wilson_score",
+          "n": 19,
+          "proportion": 0,
+          "successes": 0,
+          "upper95": 0.16817922405261626
+        }
+      }
+    },
+    "cellEvidenceCount": 80,
+    "model": "qwen/qwen3-32b",
+    "offlineExportProviderRequests": 0,
+    "primaryComparisonU1VsR": {
+      "aOnly": 0,
+      "absoluteDifference": -0.47368421052631576,
+      "absolutePairedDifference": 9,
+      "armA": "U1",
+      "armAPasses": 0,
+      "armB": "R",
+      "armBPasses": 9,
+      "bOnly": 9,
+      "both": 0,
+      "discordant": 9,
+      "exactTwoSidedPValue": 0.00390625,
+      "method": "exact_binomial_mcnemar",
+      "n": 19,
+      "neither": 10,
+      "pairedDiffCI": {
+        "difference": -0.47368421052631576,
+        "lower95": -0.47368421052631576,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": -0.19028363833937928
+      },
+      "pairedN": 19,
+      "rOnly": 0,
+      "uOnly": 9,
+      "wilsonA": {
+        "lower95": 0,
+        "method": "wilson_score",
+        "n": 19,
+        "proportion": 0,
+        "successes": 0,
+        "upper95": 0.16817922405261626
+      },
+      "wilsonB": {
+        "lower95": 0.2732980516117956,
+        "method": "wilson_score",
+        "n": 19,
+        "proportion": 0.47368421052631576,
+        "successes": 9,
+        "upper95": 0.6829219075488682
+      }
+    },
+    "primaryDenominator": 19,
+    "protocolId": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+    "providerLabel": "openrouter",
+    "recordKind": "PUBLIC_CALIBRATION_AGGREGATE_RESULTS",
+    "schemaVersion": 1,
+    "sentinelExcludedFromPrimary": true,
+    "sentinelTaskId": "cal-l10",
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "upstreamLabel": "deepinfra"
+  },
+  {
+    "cellEvidenceCount": 152,
+    "deepseekPrivacyGuardrail404Cells": 38,
+    "expectedCellEvidenceCount": 152,
+    "frozenDecisionLabels": {
+      "CROSS_MODEL_GENERALIZATION_SIGNAL": false,
+      "MODEL_STRENGTH_INTERACTION_DIRECTIONAL": {
+        "descriptiveOnly": true,
+        "notAParameterScalingLaw": true,
+        "note": "Reported as a descriptive cross-model pattern only; provider and model-family differences are confounded.",
+        "value": true
+      },
+      "READY_FOR_PUBLICATION_V0_1": true
+    },
+    "infrastructureValidCells": 113,
+    "infrastructureValidRate": 0.743421052631579,
+    "latencySummary": {
+      "maxMs": 901820,
+      "meanMs": 34039.84868421053,
+      "medianMs": 10686,
+      "minMs": 25,
+      "n": 152
+    },
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "offlineExportProviderRequests": 0,
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerRequestsDuringExport": 0,
+    "publicConclusion": "Interface advantage replicated for Qwen3 14B and Qwen3 235B A22B; it did not replicate for Qwen3 8B, DeepSeek V4 Pro was unassessable due to privacy-guardrail 404 failures, and the registered cross-model generalization signal is false.",
+    "publicationReadiness": {
+      "READY_FOR_PUBLICATION_V0_1": true,
+      "claimsHonest": true,
+      "leakagePending": false,
+      "noFalseCrossModelGeneralizationClaim": true,
+      "note": "Public export leakage scan passed; internal post-run readiness remains conditioned on complete evidence and honest claims.",
+      "publicExportLeakagePass": true,
+      "studyComplete": true
+    },
+    "qwen32bExternalAnchor": {
+      "anchorLabel": "EXISTING_QWEN3_32B_CALIBRATION_ANCHOR_NOT_REEXECUTED",
+      "importedCellCount": 38,
+      "notInPaidDenominator": true,
+      "proofSha256": "6d9da979bebdf845c9dca44ebc0966e119ee09d5c8bfd527e5860f610f78b989",
+      "sourceEvidenceTip": "c8e876133ed2745f23bc14d00aca79d68e83d590"
+    },
+    "recordKind": "PUBLIC_MULTIMODEL_LIVE_AGGREGATE_RESULTS",
+    "schemaVersion": 1,
+    "studyComplete": true,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "tokenAndCostSummary": {
+      "completionTokens": 266900,
+      "confirmedCostUsd": 0.23197088399999993,
+      "conservativeAccountingCostUsd": 1.1398656208421047,
+      "promptTokens": 115810,
+      "reservationFallbackCostUsd": 0.9078947368421049,
+      "totalTokens": 382710
+    },
+    "transportConstructedDuringExport": false
+  },
+  {
+    "cellCount": 38,
+    "costSummary": {
+      "cellsChargedByReservationFallback": 1,
+      "cellsWithConfirmedCost": 37,
+      "completionTokens": 135942,
+      "confirmedCostUsd": 0.06627527399999998,
+      "conservativeAccountingCostUsd": 0.07417001084210526,
+      "liveRequestCount": 38,
+      "promptTokens": 37792,
+      "reservationFallbackCostUsd": 0.007894736842105263,
+      "totalTokens": 173734
+    },
+    "deepseekPrivacyGuardrail404Count": 0,
+    "displayName": "Qwen3 8B",
+    "frozenDecisionLabels": {
+      "INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL": false,
+      "NORMALIZATION_MATERIALLY_RECOVERS_MODEL": false
+    },
+    "httpStatusClassCounts": {
+      "200": 38
+    },
+    "infrastructureValidCells": 37,
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "qwen/qwen3-8b",
+    "modelKey": "qwen3-8b",
+    "nativeStrict": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0.0922176518522127,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0.18421052631578946,
+        "successes": 7,
+        "upper95": 0.334188580711314
+      },
+      "hiddenPassRate": 0.18421052631578946,
+      "hiddenPasses": 7,
+      "pairedDifferenceCI": {
+        "difference": 0.16666666666666666,
+        "lower95": -0.1097176485746294,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": 0.3249408368857338
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0.16666666666666669,
+        "both": 0,
+        "discordant": 7,
+        "exactTwoSidedPValue": 0.453125,
+        "method": "exact_binomial_mcnemar",
+        "n": 18,
+        "neither": 11,
+        "rOnly": 5,
+        "uOnly": 2
+      },
+      "policy": "native-strict",
+      "visiblePasses": 7
+    },
+    "offlineExportProviderRequests": 0,
+    "oneOuterFence": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0.0922176518522127,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0.18421052631578946,
+        "successes": 7,
+        "upper95": 0.334188580711314
+      },
+      "hiddenPassRate": 0.18421052631578946,
+      "hiddenPasses": 7,
+      "pairedDifferenceCI": {
+        "difference": 0.16666666666666666,
+        "lower95": -0.1097176485746294,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": 0.3249408368857338
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0.16666666666666669,
+        "both": 0,
+        "discordant": 7,
+        "exactTwoSidedPValue": 0.453125,
+        "method": "exact_binomial_mcnemar",
+        "n": 18,
+        "neither": 11,
+        "rOnly": 5,
+        "uOnly": 2
+      },
+      "policy": "one-outer-fence",
+      "visiblePasses": 7
+    },
+    "parameterization": "dense 8.2B",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS",
+    "recoveredCells": 0,
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "terminalClassificationCounts": {
+      "MODEL_EDIT_APPLIED_ALL_EVALUATIONS_PASS": 7,
+      "MODEL_EDIT_APPLIED_VISIBLE_FAIL": 2,
+      "MODEL_EDIT_PATH_INVALID": 10,
+      "MODEL_EDIT_TARGET_AMBIGUOUS": 1,
+      "MODEL_EMITTED_INVALID_ENVELOPE": 17,
+      "PROVIDER_TRANSPORT_FAILURE": 1
+    },
+    "transportConstructedDuringExport": false,
+    "transportRequestFailures": 0
+  },
+  {
+    "cellCount": 38,
+    "costSummary": {
+      "cellsChargedByReservationFallback": 0,
+      "cellsWithConfirmedCost": 38,
+      "completionTokens": 60172,
+      "confirmedCostUsd": 0.019124640000000002,
+      "conservativeAccountingCostUsd": 0.019124640000000002,
+      "liveRequestCount": 38,
+      "promptTokens": 39028,
+      "reservationFallbackCostUsd": 0,
+      "totalTokens": 99200
+    },
+    "deepseekPrivacyGuardrail404Count": 0,
+    "displayName": "Qwen3 14B",
+    "frozenDecisionLabels": {
+      "INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL": true,
+      "NORMALIZATION_MATERIALLY_RECOVERS_MODEL": false
+    },
+    "httpStatusClassCounts": {
+      "200": 38
+    },
+    "infrastructureValidCells": 38,
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "qwen/qwen3-14b",
+    "modelKey": "qwen3-14b",
+    "nativeStrict": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0.23384219247155957,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0.3684210526315789,
+        "successes": 14,
+        "upper95": 0.5271604044299099
+      },
+      "hiddenPassRate": 0.3684210526315789,
+      "hiddenPasses": 14,
+      "pairedDifferenceCI": {
+        "difference": 0.631578947368421,
+        "lower95": 0.27309277690724687,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": 0.7180935775294239
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0.631578947368421,
+        "both": 0,
+        "discordant": 14,
+        "exactTwoSidedPValue": 0.0018310546875,
+        "method": "exact_binomial_mcnemar",
+        "n": 19,
+        "neither": 5,
+        "rOnly": 13,
+        "uOnly": 1
+      },
+      "policy": "native-strict",
+      "visiblePasses": 14
+    },
+    "offlineExportProviderRequests": 0,
+    "oneOuterFence": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0.23384219247155957,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0.3684210526315789,
+        "successes": 14,
+        "upper95": 0.5271604044299099
+      },
+      "hiddenPassRate": 0.3684210526315789,
+      "hiddenPasses": 14,
+      "pairedDifferenceCI": {
+        "difference": 0.631578947368421,
+        "lower95": 0.27309277690724687,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": 0.7180935775294239
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0.631578947368421,
+        "both": 0,
+        "discordant": 14,
+        "exactTwoSidedPValue": 0.0018310546875,
+        "method": "exact_binomial_mcnemar",
+        "n": 19,
+        "neither": 5,
+        "rOnly": 13,
+        "uOnly": 1
+      },
+      "policy": "one-outer-fence",
+      "visiblePasses": 14
+    },
+    "parameterization": "dense 14.8B",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS",
+    "recoveredCells": 0,
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "terminalClassificationCounts": {
+      "MODEL_EDIT_APPLIED_ALL_EVALUATIONS_PASS": 14,
+      "MODEL_EDIT_APPLIED_VISIBLE_FAIL": 2,
+      "MODEL_EDIT_NOT_APPLICABLE": 2,
+      "MODEL_EDIT_TARGET_AMBIGUOUS": 2,
+      "MODEL_EMITTED_INVALID_ENVELOPE": 18
+    },
+    "transportConstructedDuringExport": false,
+    "transportRequestFailures": 0
+  },
+  {
+    "cellCount": 38,
+    "costSummary": {
+      "cellsChargedByReservationFallback": 0,
+      "cellsWithConfirmedCost": 38,
+      "completionTokens": 70786,
+      "confirmedCostUsd": 0.14657096999999997,
+      "conservativeAccountingCostUsd": 0.14657096999999997,
+      "liveRequestCount": 38,
+      "promptTokens": 38990,
+      "reservationFallbackCostUsd": 0,
+      "totalTokens": 109776
+    },
+    "deepseekPrivacyGuardrail404Count": 0,
+    "displayName": "Qwen3 235B A22B",
+    "frozenDecisionLabels": {
+      "INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL": true,
+      "NORMALIZATION_MATERIALLY_RECOVERS_MODEL": false
+    },
+    "httpStatusClassCounts": {
+      "200": 38
+    },
+    "infrastructureValidCells": 38,
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "qwen/qwen3-235b-a22b",
+    "modelKey": "qwen3-235b-a22b",
+    "nativeStrict": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0.34849928364395105,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0.5,
+        "successes": 19,
+        "upper95": 0.6515007163560489
+      },
+      "hiddenPassRate": 0.5,
+      "hiddenPasses": 19,
+      "pairedDifferenceCI": {
+        "difference": 0.6842105263157895,
+        "lower95": 0.3720800167784652,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": 0.6842105263157895
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0.6842105263157894,
+        "both": 3,
+        "discordant": 13,
+        "exactTwoSidedPValue": 0.000244140625,
+        "method": "exact_binomial_mcnemar",
+        "n": 19,
+        "neither": 3,
+        "rOnly": 13,
+        "uOnly": 0
+      },
+      "policy": "native-strict",
+      "visiblePasses": 19
+    },
+    "offlineExportProviderRequests": 0,
+    "oneOuterFence": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0.34849928364395105,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0.5,
+        "successes": 19,
+        "upper95": 0.6515007163560489
+      },
+      "hiddenPassRate": 0.5,
+      "hiddenPasses": 19,
+      "pairedDifferenceCI": {
+        "difference": 0.6842105263157895,
+        "lower95": 0.3720800167784652,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.",
+        "upper95": 0.6842105263157895
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0.6842105263157894,
+        "both": 3,
+        "discordant": 13,
+        "exactTwoSidedPValue": 0.000244140625,
+        "method": "exact_binomial_mcnemar",
+        "n": 19,
+        "neither": 3,
+        "rOnly": 13,
+        "uOnly": 0
+      },
+      "policy": "one-outer-fence",
+      "visiblePasses": 19
+    },
+    "parameterization": "MoE 235B total / 22B active",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS",
+    "recoveredCells": 0,
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "terminalClassificationCounts": {
+      "MODEL_EDIT_APPLIED_ALL_EVALUATIONS_PASS": 19,
+      "MODEL_EDIT_APPLIED_VISIBLE_FAIL": 3,
+      "MODEL_EDIT_NOT_APPLICABLE": 1,
+      "MODEL_EDIT_TARGET_AMBIGUOUS": 1,
+      "MODEL_EMITTED_INVALID_ENVELOPE": 14
+    },
+    "transportConstructedDuringExport": false,
+    "transportRequestFailures": 0
+  },
+  {
+    "cellCount": 38,
+    "costSummary": {
+      "cellsChargedByReservationFallback": 38,
+      "cellsWithConfirmedCost": 0,
+      "completionTokens": 0,
+      "confirmedCostUsd": 0,
+      "conservativeAccountingCostUsd": 0.8999999999999996,
+      "liveRequestCount": 38,
+      "promptTokens": 0,
+      "reservationFallbackCostUsd": 0.8999999999999996,
+      "totalTokens": 0
+    },
+    "deepseekPrivacyGuardrail404Count": 38,
+    "displayName": "DeepSeek V4 Pro",
+    "frozenDecisionLabels": {
+      "INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL": false,
+      "NORMALIZATION_MATERIALLY_RECOVERS_MODEL": false
+    },
+    "httpStatusClassCounts": {
+      "404": 38
+    },
+    "infrastructureValidCells": 0,
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "deepseek/deepseek-v4-pro",
+    "modelKey": "deepseek-v4-pro",
+    "nativeStrict": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0,
+        "successes": 0,
+        "upper95": 0.09180986822558396
+      },
+      "hiddenPassRate": 0,
+      "hiddenPasses": 0,
+      "pairedDifferenceCI": {
+        "difference": 0,
+        "lower95": 0,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "No discordant pairs; CI collapses to the point estimate.",
+        "upper95": 0
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0,
+        "both": 0,
+        "discordant": 0,
+        "exactTwoSidedPValue": 1,
+        "method": "exact_binomial_mcnemar",
+        "n": 0,
+        "neither": 0,
+        "rOnly": 0,
+        "uOnly": 0
+      },
+      "policy": "native-strict",
+      "visiblePasses": 0
+    },
+    "offlineExportProviderRequests": 0,
+    "oneOuterFence": {
+      "cellCount": 38,
+      "hiddenPassCI": {
+        "lower95": 0,
+        "method": "wilson_score",
+        "n": 38,
+        "proportion": 0,
+        "successes": 0,
+        "upper95": 0.09180986822558396
+      },
+      "hiddenPassRate": 0,
+      "hiddenPasses": 0,
+      "pairedDifferenceCI": {
+        "difference": 0,
+        "lower95": 0,
+        "method": "newcombe_wilson_score_for_paired_difference_via_discordant_binomial",
+        "note": "No discordant pairs; CI collapses to the point estimate.",
+        "upper95": 0
+      },
+      "pairedMcnemar": {
+        "absoluteDifference": 0,
+        "both": 0,
+        "discordant": 0,
+        "exactTwoSidedPValue": 1,
+        "method": "exact_binomial_mcnemar",
+        "n": 0,
+        "neither": 0,
+        "rOnly": 0,
+        "uOnly": 0
+      },
+      "policy": "one-outer-fence",
+      "visiblePasses": 0
+    },
+    "parameterization": "MoE 1.6T total / 49B active",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS",
+    "recoveredCells": 0,
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "terminalClassificationCounts": {
+      "PROVIDER_REQUEST_FAILURE": 38
+    },
+    "transportConstructedDuringExport": false,
+    "transportRequestFailures": 38
+  }
+]

data/aggregate_results.jsonl

@@ -0,0 +1,6 @@
+{"arms":{"R":{"applicable":9,"fenceRemoved":0,"hiddenPasses":9,"lfAppended":0,"n":19,"validEnvelopes":11,"visiblePasses":9,"wilsonHidden":{"lower95":0.2732980516117956,"method":"wilson_score","n":19,"proportion":0.47368421052631576,"successes":9,"upper95":0.6829219075488682}},"RN":{"applicable":15,"fenceRemoved":11,"hiddenPasses":14,"lfAppended":0,"n":19,"validEnvelopes":18,"visiblePasses":14,"wilsonHidden":{"lower95":0.5120844910218489,"method":"wilson_score","n":19,"proportion":0.7368421052631579,"successes":14,"upper95":0.8819358765321749}},"U0":{"applicable":0,"fenceRemoved":0,"hiddenPasses":0,"lfAppended":0,"n":19,"validEnvelopes":0,"visiblePasses":0,"wilsonHidden":{"lower95":0,"method":"wilson_score","n":19,"proportion":0,"successes":0,"upper95":0.16817922405261626}},"U1":{"applicable":0,"fenceRemoved":0,"hiddenPasses":0,"lfAppended":1,"n":19,"validEnvelopes":0,"visiblePasses":0,"wilsonHidden":{"lower95":0,"method":"wilson_score","n":19,"proportion":0,"successes":0,"upper95":0.16817922405261626}}},"cellEvidenceCount":80,"model":"qwen/qwen3-32b","offlineExportProviderRequests":0,"primaryComparisonU1VsR":{"aOnly":0,"absoluteDifference":-0.47368421052631576,"absolutePairedDifference":9,"armA":"U1","armAPasses":0,"armB":"R","armBPasses":9,"bOnly":9,"both":0,"discordant":9,"exactTwoSidedPValue":0.00390625,"method":"exact_binomial_mcnemar","n":19,"neither":10,"pairedDiffCI":{"difference":-0.47368421052631576,"lower95":-0.47368421052631576,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":-0.19028363833937928},"pairedN":19,"rOnly":0,"uOnly":9,"wilsonA":{"lower95":0,"method":"wilson_score","n":19,"proportion":0,"successes":0,"upper95":0.16817922405261626},"wilsonB":{"lower95":0.2732980516117956,"method":"wilson_score","n":19,"proportion":0.47368421052631576,"successes":9,"upper95":0.6829219075488682}},"primaryDenominator":19,"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","providerLabel":"openrouter","recordKind":"PUBLIC_CALIBRATION_AGGREGATE_RESULTS","schemaVersion":1,"sentinelExcludedFromPrimary":true,"sentinelTaskId":"cal-l10","studyId":"workledger-structured-edit-interface-study-v0.1","upstreamLabel":"deepinfra"}
+{"cellEvidenceCount":152,"deepseekPrivacyGuardrail404Cells":38,"expectedCellEvidenceCount":152,"frozenDecisionLabels":{"CROSS_MODEL_GENERALIZATION_SIGNAL":false,"MODEL_STRENGTH_INTERACTION_DIRECTIONAL":{"descriptiveOnly":true,"notAParameterScalingLaw":true,"note":"Reported as a descriptive cross-model pattern only; provider and model-family differences are confounded.","value":true},"READY_FOR_PUBLICATION_V0_1":true},"infrastructureValidCells":113,"infrastructureValidRate":0.743421052631579,"latencySummary":{"maxMs":901820,"meanMs":34039.84868421053,"medianMs":10686,"minMs":25,"n":152},"liveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","offlineExportProviderRequests":0,"protocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","providerRequestsDuringExport":0,"publicConclusion":"Interface advantage replicated for Qwen3 14B and Qwen3 235B A22B; it did not replicate for Qwen3 8B, DeepSeek V4 Pro was unassessable due to privacy-guardrail 404 failures, and the registered cross-model generalization signal is false.","publicationReadiness":{"READY_FOR_PUBLICATION_V0_1":true,"claimsHonest":true,"leakagePending":false,"noFalseCrossModelGeneralizationClaim":true,"note":"Public export leakage scan passed; internal post-run readiness remains conditioned on complete evidence and honest claims.","publicExportLeakagePass":true,"studyComplete":true},"qwen32bExternalAnchor":{"anchorLabel":"EXISTING_QWEN3_32B_CALIBRATION_ANCHOR_NOT_REEXECUTED","importedCellCount":38,"notInPaidDenominator":true,"proofSha256":"6d9da979bebdf845c9dca44ebc0966e119ee09d5c8bfd527e5860f610f78b989","sourceEvidenceTip":"c8e876133ed2745f23bc14d00aca79d68e83d590"},"recordKind":"PUBLIC_MULTIMODEL_LIVE_AGGREGATE_RESULTS","schemaVersion":1,"studyComplete":true,"studyId":"workledger-structured-edit-interface-study-v0.1","tokenAndCostSummary":{"completionTokens":266900,"confirmedCostUsd":0.23197088399999993,"conservativeAccountingCostUsd":1.1398656208421047,"promptTokens":115810,"reservationFallbackCostUsd":0.9078947368421049,"totalTokens":382710},"transportConstructedDuringExport":false}
+{"cellCount":38,"costSummary":{"cellsChargedByReservationFallback":1,"cellsWithConfirmedCost":37,"completionTokens":135942,"confirmedCostUsd":0.06627527399999998,"conservativeAccountingCostUsd":0.07417001084210526,"liveRequestCount":38,"promptTokens":37792,"reservationFallbackCostUsd":0.007894736842105263,"totalTokens":173734},"deepseekPrivacyGuardrail404Count":0,"displayName":"Qwen3 8B","frozenDecisionLabels":{"INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL":false,"NORMALIZATION_MATERIALLY_RECOVERS_MODEL":false},"httpStatusClassCounts":{"200":38},"infrastructureValidCells":37,"liveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","modelId":"qwen/qwen3-8b","modelKey":"qwen3-8b","nativeStrict":{"cellCount":38,"hiddenPassCI":{"lower95":0.0922176518522127,"method":"wilson_score","n":38,"proportion":0.18421052631578946,"successes":7,"upper95":0.334188580711314},"hiddenPassRate":0.18421052631578946,"hiddenPasses":7,"pairedDifferenceCI":{"difference":0.16666666666666666,"lower95":-0.1097176485746294,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":0.3249408368857338},"pairedMcnemar":{"absoluteDifference":0.16666666666666669,"both":0,"discordant":7,"exactTwoSidedPValue":0.453125,"method":"exact_binomial_mcnemar","n":18,"neither":11,"rOnly":5,"uOnly":2},"policy":"native-strict","visiblePasses":7},"offlineExportProviderRequests":0,"oneOuterFence":{"cellCount":38,"hiddenPassCI":{"lower95":0.0922176518522127,"method":"wilson_score","n":38,"proportion":0.18421052631578946,"successes":7,"upper95":0.334188580711314},"hiddenPassRate":0.18421052631578946,"hiddenPasses":7,"pairedDifferenceCI":{"difference":0.16666666666666666,"lower95":-0.1097176485746294,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":0.3249408368857338},"pairedMcnemar":{"absoluteDifference":0.16666666666666669,"both":0,"discordant":7,"exactTwoSidedPValue":0.453125,"method":"exact_binomial_mcnemar","n":18,"neither":11,"rOnly":5,"uOnly":2},"policy":"one-outer-fence","visiblePasses":7},"parameterization":"dense 8.2B","protocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","providerRequestsDuringExport":0,"recordKind":"PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS","recoveredCells":0,"schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1","terminalClassificationCounts":{"MODEL_EDIT_APPLIED_ALL_EVALUATIONS_PASS":7,"MODEL_EDIT_APPLIED_VISIBLE_FAIL":2,"MODEL_EDIT_PATH_INVALID":10,"MODEL_EDIT_TARGET_AMBIGUOUS":1,"MODEL_EMITTED_INVALID_ENVELOPE":17,"PROVIDER_TRANSPORT_FAILURE":1},"transportConstructedDuringExport":false,"transportRequestFailures":0}
+{"cellCount":38,"costSummary":{"cellsChargedByReservationFallback":0,"cellsWithConfirmedCost":38,"completionTokens":60172,"confirmedCostUsd":0.019124640000000002,"conservativeAccountingCostUsd":0.019124640000000002,"liveRequestCount":38,"promptTokens":39028,"reservationFallbackCostUsd":0,"totalTokens":99200},"deepseekPrivacyGuardrail404Count":0,"displayName":"Qwen3 14B","frozenDecisionLabels":{"INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL":true,"NORMALIZATION_MATERIALLY_RECOVERS_MODEL":false},"httpStatusClassCounts":{"200":38},"infrastructureValidCells":38,"liveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","modelId":"qwen/qwen3-14b","modelKey":"qwen3-14b","nativeStrict":{"cellCount":38,"hiddenPassCI":{"lower95":0.23384219247155957,"method":"wilson_score","n":38,"proportion":0.3684210526315789,"successes":14,"upper95":0.5271604044299099},"hiddenPassRate":0.3684210526315789,"hiddenPasses":14,"pairedDifferenceCI":{"difference":0.631578947368421,"lower95":0.27309277690724687,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":0.7180935775294239},"pairedMcnemar":{"absoluteDifference":0.631578947368421,"both":0,"discordant":14,"exactTwoSidedPValue":0.0018310546875,"method":"exact_binomial_mcnemar","n":19,"neither":5,"rOnly":13,"uOnly":1},"policy":"native-strict","visiblePasses":14},"offlineExportProviderRequests":0,"oneOuterFence":{"cellCount":38,"hiddenPassCI":{"lower95":0.23384219247155957,"method":"wilson_score","n":38,"proportion":0.3684210526315789,"successes":14,"upper95":0.5271604044299099},"hiddenPassRate":0.3684210526315789,"hiddenPasses":14,"pairedDifferenceCI":{"difference":0.631578947368421,"lower95":0.27309277690724687,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":0.7180935775294239},"pairedMcnemar":{"absoluteDifference":0.631578947368421,"both":0,"discordant":14,"exactTwoSidedPValue":0.0018310546875,"method":"exact_binomial_mcnemar","n":19,"neither":5,"rOnly":13,"uOnly":1},"policy":"one-outer-fence","visiblePasses":14},"parameterization":"dense 14.8B","protocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","providerRequestsDuringExport":0,"recordKind":"PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS","recoveredCells":0,"schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1","terminalClassificationCounts":{"MODEL_EDIT_APPLIED_ALL_EVALUATIONS_PASS":14,"MODEL_EDIT_APPLIED_VISIBLE_FAIL":2,"MODEL_EDIT_NOT_APPLICABLE":2,"MODEL_EDIT_TARGET_AMBIGUOUS":2,"MODEL_EMITTED_INVALID_ENVELOPE":18},"transportConstructedDuringExport":false,"transportRequestFailures":0}
+{"cellCount":38,"costSummary":{"cellsChargedByReservationFallback":0,"cellsWithConfirmedCost":38,"completionTokens":70786,"confirmedCostUsd":0.14657096999999997,"conservativeAccountingCostUsd":0.14657096999999997,"liveRequestCount":38,"promptTokens":38990,"reservationFallbackCostUsd":0,"totalTokens":109776},"deepseekPrivacyGuardrail404Count":0,"displayName":"Qwen3 235B A22B","frozenDecisionLabels":{"INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL":true,"NORMALIZATION_MATERIALLY_RECOVERS_MODEL":false},"httpStatusClassCounts":{"200":38},"infrastructureValidCells":38,"liveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","modelId":"qwen/qwen3-235b-a22b","modelKey":"qwen3-235b-a22b","nativeStrict":{"cellCount":38,"hiddenPassCI":{"lower95":0.34849928364395105,"method":"wilson_score","n":38,"proportion":0.5,"successes":19,"upper95":0.6515007163560489},"hiddenPassRate":0.5,"hiddenPasses":19,"pairedDifferenceCI":{"difference":0.6842105263157895,"lower95":0.3720800167784652,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":0.6842105263157895},"pairedMcnemar":{"absoluteDifference":0.6842105263157894,"both":3,"discordant":13,"exactTwoSidedPValue":0.000244140625,"method":"exact_binomial_mcnemar","n":19,"neither":3,"rOnly":13,"uOnly":0},"policy":"native-strict","visiblePasses":19},"offlineExportProviderRequests":0,"oneOuterFence":{"cellCount":38,"hiddenPassCI":{"lower95":0.34849928364395105,"method":"wilson_score","n":38,"proportion":0.5,"successes":19,"upper95":0.6515007163560489},"hiddenPassRate":0.5,"hiddenPasses":19,"pairedDifferenceCI":{"difference":0.6842105263157895,"lower95":0.3720800167784652,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"Derived from Wilson CI on P(R-only | discordant), scaled by discordant/n.","upper95":0.6842105263157895},"pairedMcnemar":{"absoluteDifference":0.6842105263157894,"both":3,"discordant":13,"exactTwoSidedPValue":0.000244140625,"method":"exact_binomial_mcnemar","n":19,"neither":3,"rOnly":13,"uOnly":0},"policy":"one-outer-fence","visiblePasses":19},"parameterization":"MoE 235B total / 22B active","protocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","providerRequestsDuringExport":0,"recordKind":"PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS","recoveredCells":0,"schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1","terminalClassificationCounts":{"MODEL_EDIT_APPLIED_ALL_EVALUATIONS_PASS":19,"MODEL_EDIT_APPLIED_VISIBLE_FAIL":3,"MODEL_EDIT_NOT_APPLICABLE":1,"MODEL_EDIT_TARGET_AMBIGUOUS":1,"MODEL_EMITTED_INVALID_ENVELOPE":14},"transportConstructedDuringExport":false,"transportRequestFailures":0}
+{"cellCount":38,"costSummary":{"cellsChargedByReservationFallback":38,"cellsWithConfirmedCost":0,"completionTokens":0,"confirmedCostUsd":0,"conservativeAccountingCostUsd":0.8999999999999996,"liveRequestCount":38,"promptTokens":0,"reservationFallbackCostUsd":0.8999999999999996,"totalTokens":0},"deepseekPrivacyGuardrail404Count":38,"displayName":"DeepSeek V4 Pro","frozenDecisionLabels":{"INTERFACE_ADVANTAGE_REPLICATES_FOR_MODEL":false,"NORMALIZATION_MATERIALLY_RECOVERS_MODEL":false},"httpStatusClassCounts":{"404":38},"infrastructureValidCells":0,"liveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","modelId":"deepseek/deepseek-v4-pro","modelKey":"deepseek-v4-pro","nativeStrict":{"cellCount":38,"hiddenPassCI":{"lower95":0,"method":"wilson_score","n":38,"proportion":0,"successes":0,"upper95":0.09180986822558396},"hiddenPassRate":0,"hiddenPasses":0,"pairedDifferenceCI":{"difference":0,"lower95":0,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"No discordant pairs; CI collapses to the point estimate.","upper95":0},"pairedMcnemar":{"absoluteDifference":0,"both":0,"discordant":0,"exactTwoSidedPValue":1,"method":"exact_binomial_mcnemar","n":0,"neither":0,"rOnly":0,"uOnly":0},"policy":"native-strict","visiblePasses":0},"offlineExportProviderRequests":0,"oneOuterFence":{"cellCount":38,"hiddenPassCI":{"lower95":0,"method":"wilson_score","n":38,"proportion":0,"successes":0,"upper95":0.09180986822558396},"hiddenPassRate":0,"hiddenPasses":0,"pairedDifferenceCI":{"difference":0,"lower95":0,"method":"newcombe_wilson_score_for_paired_difference_via_discordant_binomial","note":"No discordant pairs; CI collapses to the point estimate.","upper95":0},"pairedMcnemar":{"absoluteDifference":0,"both":0,"discordant":0,"exactTwoSidedPValue":1,"method":"exact_binomial_mcnemar","n":0,"neither":0,"rOnly":0,"uOnly":0},"policy":"one-outer-fence","visiblePasses":0},"parameterization":"MoE 1.6T total / 49B active","protocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","providerRequestsDuringExport":0,"recordKind":"PUBLIC_MULTIMODEL_MODEL_AGGREGATE_RESULTS","recoveredCells":0,"schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1","terminalClassificationCounts":{"PROVIDER_REQUEST_FAILURE":38},"transportConstructedDuringExport":false,"transportRequestFailures":38}

data/claim-survival.json

@@ -0,0 +1,252 @@
+[
+  {
+    "claim": "Strict exact_replace outperformed strict unified_diff on the frozen expansion bank.",
+    "classification": "SUPPORTED",
+    "evidenceAgainstOrLimits": [
+      "Expansion bank was replacement-affine and single-file by construction."
+    ],
+    "evidenceFor": [
+      "Expansion aggregate: R 13/30 hidden passes vs U 0/30; McNemar p=0.000244140625.",
+      "Expansion red-team replay reproduced the result and found no suspicious R passes."
+    ],
+    "id": 1,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "On the frozen replacement-affine expansion bank, strict exact_replace outperformed the frozen strict unified_diff stack.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "The pilot provided directional support for exact_replace before expansion.",
+    "classification": "SUPPORTED_WITH_NARROW_SCOPE",
+    "evidenceAgainstOrLimits": [
+      "Pilot n=5 primary tasks; no broad inference alone."
+    ],
+    "evidenceFor": [
+      "Pilot aggregate: R 2/5 primary hidden passes vs U 0/5."
+    ],
+    "id": 2,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "The pilot was directional only and motivated the larger expansion.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "The four-arm calibration result survived a fairer practical unified-diff arm.",
+    "classification": "SUPPORTED_WITH_NARROW_SCOPE",
+    "evidenceAgainstOrLimits": [
+      "One model/provider/config; single attempt; cal-l10 sentinel excluded."
+    ],
+    "evidenceFor": [
+      "Calibration primary U1 vs R: U1 0/19 hidden passes, R 9/19, p=0.00390625.",
+      "Low-affinity primary stratum: U1 0/9, R 6/9."
+    ],
+    "id": 3,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "For this Qwen3-32B calibrated bank, R remained ahead of the practical U1 arm.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "RN normalization improved operational exact_replace success.",
+    "classification": "DIRECTIONALLY_SUPPORTED",
+    "evidenceAgainstOrLimits": [
+      "R and RN were independent stochastic requests with byte-identical prompts; R-vs-RN p=0.1796875."
+    ],
+    "evidenceFor": [
+      "Calibration primary RN hidden passes 14/19 vs R 9/19.",
+      "RN removed outer fences in 11 primary cells."
+    ],
+    "id": 4,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "RN showed operational benefit in this run, but should be reported separately from native strict R compliance.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "U1 repaired the legacy U0 unified-diff failure mode.",
+    "classification": "UNSUPPORTED_AS_WORDED",
+    "evidenceAgainstOrLimits": [
+      "Calibration primary U0 hidden passes 0/19 and U1 hidden passes 0/19."
+    ],
+    "evidenceFor": [
+      "U1 added scaffolded practical instructions and LF appending."
+    ],
+    "id": 5,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "U1 did not produce hidden-pass gains for Qwen3-32B in this calibration run.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "The exact_replace advantage persists in low-affinity tasks.",
+    "classification": "SUPPORTED_WITH_NARROW_SCOPE",
+    "evidenceAgainstOrLimits": [
+      "The low-affinity stratum still contains many contiguous exact_replace gold spans."
+    ],
+    "evidenceFor": [
+      "Calibration low-affinity primary stratum: U1 0/9 hidden passes, R 6/9, p=0.03125.",
+      "This audit found cal-l01..cal-l09 gold edits expressible by both arms; cal-l10 is excluded sentinel."
+    ],
+    "id": 6,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "Within the calibrated low-affinity primary subset, R outperformed U1; this does not prove broad multi-file superiority.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "exact_replace is broadly preferable to unified_diff for agentic coding.",
+    "classification": "INSUFFICIENT_EVIDENCE",
+    "evidenceAgainstOrLimits": [
+      "No production workload, movement-heavy edits, large multi-file edits, or multi-model comparison yet.",
+      "cal-l10 shows evaluators can accept architecture-bypassing single-file alternatives."
+    ],
+    "evidenceFor": [
+      "Expansion and calibration both favor R on their registered banks."
+    ],
+    "id": 7,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "The evidence supports exact_replace for bounded replacement-shaped tasks with this model, not a broad production default.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "WorkLedger improved model performance.",
+    "classification": "OVERSTATED",
+    "evidenceAgainstOrLimits": [
+      "The harness measured outcomes and often scored fail-closed lower than permissive tooling would."
+    ],
+    "evidenceFor": [
+      "WorkLedger preserved evidence and separated interface outcomes."
+    ],
+    "id": 8,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "WorkLedger made interface-dependent failures measurable; it did not by itself improve the model.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "Fail-closed validation added scientific value.",
+    "classification": "SUPPORTED",
+    "evidenceAgainstOrLimits": [
+      "Strictness margins must be documented so fail-closed does not become hidden treatment asymmetry."
+    ],
+    "evidenceFor": [
+      "Pilot, expansion, calibration, and this audit all rely on preserved raw responses, hashes, and evaluator outputs.",
+      "PR48 cal-l10 live response could be reinterpreted only because raw bytes and terminal evidence were preserved."
+    ],
+    "id": 9,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "Fail-closed, byte-preserved evidence enabled post hoc failure attribution and claim revision.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "cal-l10 demonstrates exact_replace can solve the intended two-file architecture task.",
+    "classification": "UNSUPPORTED_AS_WORDED",
+    "evidenceAgainstOrLimits": [
+      "The passing R solution edits registry.cjs directly and bypasses registry-defaults.cjs.",
+      "The hidden evaluator does not enforce the intended defaults-module architecture."
+    ],
+    "evidenceFor": [
+      "The live R response and this audit's direct-registry alternative pass visible and hidden evaluators."
+    ],
+    "id": 10,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "cal-l10 demonstrates evaluator underconstraint: a single-file R bypass passes, while the gold architecture-preserving edit remains U-only under the frozen contract.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "cal-l10 should remain a contract-coverage sentinel.",
+    "classification": "SUPPORTED",
+    "evidenceAgainstOrLimits": [
+      "Public wording should not call cal-l10 semantically U-only."
+    ],
+    "evidenceFor": [
+      "Frozen fixture marks cal-l10 primaryAggregate=false and goldReplace=null.",
+      "This audit confirms the gold edit changes registry-defaults.cjs, outside R's target-file materializer."
+    ],
+    "id": 11,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "Keep cal-l10 as a contract-coverage/gold-edit U-only sentinel; distinguish it from semantic U-only.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "The calibrated publication is ready for multi-model progression.",
+    "classification": "SUPPORTED_WITH_NARROW_SCOPE",
+    "evidenceAgainstOrLimits": [
+      "Proceed only with revised wording, preserved frozen labels, and explicit solution-space/evaluator-underconstraint caveats."
+    ],
+    "evidenceFor": [
+      "Calibration decision classification says proceed to multi-model with calibrated arms.",
+      "This offline audit provides future-only label revision and claim wording constraints."
+    ],
+    "id": 12,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "Proceed to multi-model as a calibrated interface study, carrying forward the audit caveats and future label taxonomy.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "The calibrated exact_replace interface advantage replicated across the live multimodel cohort.",
+    "classification": "SUPPORTED_WITH_NARROW_SCOPE",
+    "evidenceAgainstOrLimits": [
+      "Qwen3 8B did not meet the registered per-model replication rule.",
+      "DeepSeek V4 Pro was unassessable because all 38 cells hit the privacy-guardrail 404 failure class."
+    ],
+    "evidenceFor": [
+      "Qwen3 14B strict native R vs U1: R-only 13, U1-only 1, paired n=19.",
+      "Qwen3 235B A22B strict native R vs U1: R-only 13, U1-only 0, paired n=19."
+    ],
+    "id": 13,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "The interface advantage replicated for Qwen3 14B and Qwen3 235B A22B, but not for Qwen3 8B; DeepSeek V4 Pro was unassessable.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "The live multimodel run proves cross-model generalization.",
+    "classification": "UNSUPPORTED_AS_WORDED",
+    "evidenceAgainstOrLimits": [
+      "The registered CROSS_MODEL_GENERALIZATION_SIGNAL label is false.",
+      "The non-Qwen DeepSeek model was unassessable, and Qwen3 8B did not replicate."
+    ],
+    "evidenceFor": [
+      "Two Qwen-family models met the registered per-model interface-advantage replication rule."
+    ],
+    "id": 14,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "The run provides a narrower model-strength interaction signal, not a registered cross-model generalization claim.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "DeepSeek V4 Pro supports or refutes the interface advantage.",
+    "classification": "INSUFFICIENT_EVIDENCE",
+    "evidenceAgainstOrLimits": [
+      "All 38 DeepSeek V4 Pro cells were infrastructure-invalid privacy-guardrail 404 failures with empty final answers."
+    ],
+    "evidenceFor": [],
+    "id": 15,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "DeepSeek V4 Pro is unassessable in this live replication because no cells produced valid model outputs.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  },
+  {
+    "claim": "Fence normalization materially recovered multimodel outcomes.",
+    "classification": "UNSUPPORTED_AS_WORDED",
+    "evidenceAgainstOrLimits": [
+      "Recovered cells were 0 for each public multimodel row in PER_MODEL_ANALYSIS."
+    ],
+    "evidenceFor": [
+      "Both strict native and one-outer-fence policy outcomes are reported on the same response bytes."
+    ],
+    "id": 16,
+    "recordKind": "PUBLIC_CLAIM_SURVIVAL_RECORD",
+    "safestPublicWording": "Dual strict/fence-normalized policies were evaluated on identical bytes; they did not materially recover outcomes in this run.",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1"
+  }
+]

data/claim_survival.jsonl

@@ -0,0 +1,16 @@
+{"claim":"Strict exact_replace outperformed strict unified_diff on the frozen expansion bank.","classification":"SUPPORTED","evidenceAgainstOrLimits":["Expansion bank was replacement-affine and single-file by construction."],"evidenceFor":["Expansion aggregate: R 13/30 hidden passes vs U 0/30; McNemar p=0.000244140625.","Expansion red-team replay reproduced the result and found no suspicious R passes."],"id":1,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"On the frozen replacement-affine expansion bank, strict exact_replace outperformed the frozen strict unified_diff stack.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"The pilot provided directional support for exact_replace before expansion.","classification":"SUPPORTED_WITH_NARROW_SCOPE","evidenceAgainstOrLimits":["Pilot n=5 primary tasks; no broad inference alone."],"evidenceFor":["Pilot aggregate: R 2/5 primary hidden passes vs U 0/5."],"id":2,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"The pilot was directional only and motivated the larger expansion.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"The four-arm calibration result survived a fairer practical unified-diff arm.","classification":"SUPPORTED_WITH_NARROW_SCOPE","evidenceAgainstOrLimits":["One model/provider/config; single attempt; cal-l10 sentinel excluded."],"evidenceFor":["Calibration primary U1 vs R: U1 0/19 hidden passes, R 9/19, p=0.00390625.","Low-affinity primary stratum: U1 0/9, R 6/9."],"id":3,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"For this Qwen3-32B calibrated bank, R remained ahead of the practical U1 arm.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"RN normalization improved operational exact_replace success.","classification":"DIRECTIONALLY_SUPPORTED","evidenceAgainstOrLimits":["R and RN were independent stochastic requests with byte-identical prompts; R-vs-RN p=0.1796875."],"evidenceFor":["Calibration primary RN hidden passes 14/19 vs R 9/19.","RN removed outer fences in 11 primary cells."],"id":4,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"RN showed operational benefit in this run, but should be reported separately from native strict R compliance.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"U1 repaired the legacy U0 unified-diff failure mode.","classification":"UNSUPPORTED_AS_WORDED","evidenceAgainstOrLimits":["Calibration primary U0 hidden passes 0/19 and U1 hidden passes 0/19."],"evidenceFor":["U1 added scaffolded practical instructions and LF appending."],"id":5,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"U1 did not produce hidden-pass gains for Qwen3-32B in this calibration run.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"The exact_replace advantage persists in low-affinity tasks.","classification":"SUPPORTED_WITH_NARROW_SCOPE","evidenceAgainstOrLimits":["The low-affinity stratum still contains many contiguous exact_replace gold spans."],"evidenceFor":["Calibration low-affinity primary stratum: U1 0/9 hidden passes, R 6/9, p=0.03125.","This audit found cal-l01..cal-l09 gold edits expressible by both arms; cal-l10 is excluded sentinel."],"id":6,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"Within the calibrated low-affinity primary subset, R outperformed U1; this does not prove broad multi-file superiority.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"exact_replace is broadly preferable to unified_diff for agentic coding.","classification":"INSUFFICIENT_EVIDENCE","evidenceAgainstOrLimits":["No production workload, movement-heavy edits, large multi-file edits, or multi-model comparison yet.","cal-l10 shows evaluators can accept architecture-bypassing single-file alternatives."],"evidenceFor":["Expansion and calibration both favor R on their registered banks."],"id":7,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"The evidence supports exact_replace for bounded replacement-shaped tasks with this model, not a broad production default.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"WorkLedger improved model performance.","classification":"OVERSTATED","evidenceAgainstOrLimits":["The harness measured outcomes and often scored fail-closed lower than permissive tooling would."],"evidenceFor":["WorkLedger preserved evidence and separated interface outcomes."],"id":8,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"WorkLedger made interface-dependent failures measurable; it did not by itself improve the model.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"Fail-closed validation added scientific value.","classification":"SUPPORTED","evidenceAgainstOrLimits":["Strictness margins must be documented so fail-closed does not become hidden treatment asymmetry."],"evidenceFor":["Pilot, expansion, calibration, and this audit all rely on preserved raw responses, hashes, and evaluator outputs.","PR48 cal-l10 live response could be reinterpreted only because raw bytes and terminal evidence were preserved."],"id":9,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"Fail-closed, byte-preserved evidence enabled post hoc failure attribution and claim revision.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"cal-l10 demonstrates exact_replace can solve the intended two-file architecture task.","classification":"UNSUPPORTED_AS_WORDED","evidenceAgainstOrLimits":["The passing R solution edits registry.cjs directly and bypasses registry-defaults.cjs.","The hidden evaluator does not enforce the intended defaults-module architecture."],"evidenceFor":["The live R response and this audit's direct-registry alternative pass visible and hidden evaluators."],"id":10,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"cal-l10 demonstrates evaluator underconstraint: a single-file R bypass passes, while the gold architecture-preserving edit remains U-only under the frozen contract.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"cal-l10 should remain a contract-coverage sentinel.","classification":"SUPPORTED","evidenceAgainstOrLimits":["Public wording should not call cal-l10 semantically U-only."],"evidenceFor":["Frozen fixture marks cal-l10 primaryAggregate=false and goldReplace=null.","This audit confirms the gold edit changes registry-defaults.cjs, outside R's target-file materializer."],"id":11,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"Keep cal-l10 as a contract-coverage/gold-edit U-only sentinel; distinguish it from semantic U-only.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"The calibrated publication is ready for multi-model progression.","classification":"SUPPORTED_WITH_NARROW_SCOPE","evidenceAgainstOrLimits":["Proceed only with revised wording, preserved frozen labels, and explicit solution-space/evaluator-underconstraint caveats."],"evidenceFor":["Calibration decision classification says proceed to multi-model with calibrated arms.","This offline audit provides future-only label revision and claim wording constraints."],"id":12,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"Proceed to multi-model as a calibrated interface study, carrying forward the audit caveats and future label taxonomy.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"The calibrated exact_replace interface advantage replicated across the live multimodel cohort.","classification":"SUPPORTED_WITH_NARROW_SCOPE","evidenceAgainstOrLimits":["Qwen3 8B did not meet the registered per-model replication rule.","DeepSeek V4 Pro was unassessable because all 38 cells hit the privacy-guardrail 404 failure class."],"evidenceFor":["Qwen3 14B strict native R vs U1: R-only 13, U1-only 1, paired n=19.","Qwen3 235B A22B strict native R vs U1: R-only 13, U1-only 0, paired n=19."],"id":13,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"The interface advantage replicated for Qwen3 14B and Qwen3 235B A22B, but not for Qwen3 8B; DeepSeek V4 Pro was unassessable.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"The live multimodel run proves cross-model generalization.","classification":"UNSUPPORTED_AS_WORDED","evidenceAgainstOrLimits":["The registered CROSS_MODEL_GENERALIZATION_SIGNAL label is false.","The non-Qwen DeepSeek model was unassessable, and Qwen3 8B did not replicate."],"evidenceFor":["Two Qwen-family models met the registered per-model interface-advantage replication rule."],"id":14,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"The run provides a narrower model-strength interaction signal, not a registered cross-model generalization claim.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"DeepSeek V4 Pro supports or refutes the interface advantage.","classification":"INSUFFICIENT_EVIDENCE","evidenceAgainstOrLimits":["All 38 DeepSeek V4 Pro cells were infrastructure-invalid privacy-guardrail 404 failures with empty final answers."],"evidenceFor":[],"id":15,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"DeepSeek V4 Pro is unassessable in this live replication because no cells produced valid model outputs.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}
+{"claim":"Fence normalization materially recovered multimodel outcomes.","classification":"UNSUPPORTED_AS_WORDED","evidenceAgainstOrLimits":["Recovered cells were 0 for each public multimodel row in PER_MODEL_ANALYSIS."],"evidenceFor":["Both strict native and one-outer-fence policy outcomes are reported on the same response bytes."],"id":16,"recordKind":"PUBLIC_CLAIM_SURVIVAL_RECORD","safestPublicWording":"Dual strict/fence-normalized policies were evaluated on identical bytes; they did not materially recover outcomes in this run.","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1"}

data/model-metadata.json

@@ -0,0 +1,70 @@
+[
+  {
+    "model": "qwen/qwen3-32b",
+    "note": "Public package metadata only; private provider request/response identifiers and request bodies are withheld.",
+    "protocolId": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+    "providerLabel": "openrouter",
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MODEL_METADATA",
+    "routePinned": true,
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "upstreamLabel": "deepinfra"
+  },
+  {
+    "displayName": "Qwen3 8B",
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "qwen/qwen3-8b",
+    "note": "Provider route, endpoint, request body, response identifiers, and reasoning fields are withheld from this public metadata.",
+    "parameterization": "dense 8.2B",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerFreePublicMetadata": true,
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_METADATA",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "transportConstructedDuringExport": false
+  },
+  {
+    "displayName": "Qwen3 14B",
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "qwen/qwen3-14b",
+    "note": "Provider route, endpoint, request body, response identifiers, and reasoning fields are withheld from this public metadata.",
+    "parameterization": "dense 14.8B",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerFreePublicMetadata": true,
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_METADATA",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "transportConstructedDuringExport": false
+  },
+  {
+    "displayName": "Qwen3 235B A22B",
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "qwen/qwen3-235b-a22b",
+    "note": "Provider route, endpoint, request body, response identifiers, and reasoning fields are withheld from this public metadata.",
+    "parameterization": "MoE 235B total / 22B active",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerFreePublicMetadata": true,
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_METADATA",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "transportConstructedDuringExport": false
+  },
+  {
+    "displayName": "DeepSeek V4 Pro",
+    "liveProtocolId": "qwen3-32b-v7-calibrated-multimodel-live-v1",
+    "modelId": "deepseek/deepseek-v4-pro",
+    "note": "Provider route, endpoint, request body, response identifiers, and reasoning fields are withheld from this public metadata.",
+    "parameterization": "MoE 1.6T total / 49B active",
+    "protocolId": "qwen3-32b-v7-calibrated-multimodel-offline-v1",
+    "providerFreePublicMetadata": true,
+    "providerRequestsDuringExport": 0,
+    "recordKind": "PUBLIC_MULTIMODEL_MODEL_METADATA",
+    "schemaVersion": 1,
+    "studyId": "workledger-structured-edit-interface-study-v0.1",
+    "transportConstructedDuringExport": false
+  }
+]

data/studies.jsonl

@@ -0,0 +1 @@
+{"excludedPrivateMaterials":["hidden evaluator source","reasoning traces","provider request and response identifiers","authorization headers","idempotency keys","request bodies","raw provider response bodies","SQLite ledgers"],"exportedEvidence":"PR48 calibration final assistant answers plus sanitized four-model live replication generations, aggregates, claims, and metadata","model":"qwen/qwen3-32b","multimodelCellEvidenceCount":152,"multimodelLiveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","multimodelProtocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","offlineOnly":true,"optionalPriorStudyScope":"Pilot and expansion are represented only through public claim wording metadata, not full task or generation exports.","protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","providerLabel":"openrouter","providerRequestsDuringExport":0,"recordKind":"PUBLIC_STUDY_METADATA","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1","suggestedTag":"study/structured-edit-interface-v0.1","title":"WorkLedger structured edit interface study v0.1","transportConstructedDuringExport":false,"upstreamLabel":"deepinfra"}

data/study_metadata.jsonl

@@ -0,0 +1 @@
+{"excludedPrivateMaterials":["hidden evaluator source","reasoning traces","provider request and response identifiers","authorization headers","idempotency keys","request bodies","raw provider response bodies","SQLite ledgers"],"exportedEvidence":"PR48 calibration final assistant answers plus sanitized four-model live replication generations, aggregates, claims, and metadata","model":"qwen/qwen3-32b","multimodelCellEvidenceCount":152,"multimodelLiveProtocolId":"qwen3-32b-v7-calibrated-multimodel-live-v1","multimodelProtocolId":"qwen3-32b-v7-calibrated-multimodel-offline-v1","offlineOnly":true,"optionalPriorStudyScope":"Pilot and expansion are represented only through public claim wording metadata, not full task or generation exports.","protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","providerLabel":"openrouter","providerRequestsDuringExport":0,"recordKind":"PUBLIC_STUDY_METADATA","schemaVersion":1,"studyId":"workledger-structured-edit-interface-study-v0.1","suggestedTag":"study/structured-edit-interface-v0.1","title":"WorkLedger structured edit interface study v0.1","transportConstructedDuringExport":false,"upstreamLabel":"deepinfra"}

data/tasks.jsonl

@@ -0,0 +1,20 @@
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"missing-dunder-method","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["ordered_set.py"],"fixedSourceSha256":"94c62b4cf2194b4d87812e00a01dd71c65a96053d767621d651b1a4e9f3c3d0e","fixtureMetadataSha256":"74d61512603e94340a8ed39df061ed2c5a1e0722abcac4d83bb37a5e51432c8c","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bs06-django-ordered-set-reversed"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"1ea4533be8b3d39cb39ee8f59ce6a648bee053391ca3c4ed96d04df887a09a2f","issueStatement":"# Bug: `OrderedSet` does not support `reversed()`\n\nCalling `reversed(ordered_set)` raises a `TypeError` because `OrderedSet`\ndoes not implement `__reversed__`.\n\nAdd a `__reversed__` method that delegates to `reversed(self.dict)`\n(Python 3.8+ dicts support `reversed()`).\n","language":"python","ordinal":0,"primaryAggregate":true,"promptHashes":{"R":"a44f1aca6653a4087360124bc54854087b57ae781574b6654c293a51134fb282","RN":"a44f1aca6653a4087360124bc54854087b57ae781574b6654c293a51134fb282","U0":"43c4e35a905e67f926b4e65aa8ed33c5590c5020e731a3bf7022b9ecab957b39","U1":"8188509872020e24aaff4af680f28313894a935154523b71e048f6e50f75d618"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["ordered_set.py"],"sourceSha256":"3b5115ab49012d1189910395591a71df5e94135834ce4038d385f498908efe9b","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h01","title":"Cal high-affinity 01: OrderedSet missing __reversed__ (GUIDED)","visibleEvaluatorSha256":"51ccfa2356fb45e798ebbd34af9c4b745182e4adbbd783c00512bc8503834843","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"equality-vs-identity-check-causing-ValueError","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["mock_compat.py"],"fixedSourceSha256":"aa63779e7a41e33a9fcbe54f4de0eb1658551e7f9a45f9e15711c771a4854a2b","fixtureMetadataSha256":"2b42e3f0920d6b4117bb16aa29bfa80f7f48b8e4db6ac45acbc4eff432f0d200","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bm09-pytest-mock-sentinel"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"6e717fab1bf0aff27122012099677d58a251ae5627d1202c4774fad41462ad06","issueStatement":"# Bug: `num_mock_patch_args` crashes when `p.new` is array-like\n\nThe function uses `p.new in sentinels` to check if a patch's `.new` value\nis the DEFAULT sentinel. For array-like objects (numpy arrays), the `in`\noperator calls `__eq__` which returns an array, and the truthiness test\nraises `ValueError: The truth value of an array is ambiguous`.\n\nFix: replace the `in` membership test with identity checks using `is`:\n```python\np.new is mock_sentinel or p.new is ut_mock_sentinel\n```\n","language":"python","ordinal":1,"primaryAggregate":true,"promptHashes":{"R":"28eae0ace1ab883edb16c125254bbf7ce6ee6b0fbabbaf89ee99b1bf3abf62bb","RN":"28eae0ace1ab883edb16c125254bbf7ce6ee6b0fbabbaf89ee99b1bf3abf62bb","U0":"493c5a312b461306370308d325b54483d0c51a7a8129c599194d912334fda466","U1":"4d7a2f84e3cd83fa60db56ab8462eec4717cf2d1befc4b987a176dd99df1cfa2"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["mock_compat.py"],"sourceSha256":"dfbeeb8605cc811b12579219819858675ee990b1962432cfdf68c98d054265a2","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h02","title":"Cal high-affinity 02: num_mock_patch_args uses `in` sentinel check (GUIDED)","visibleEvaluatorSha256":"48b1969c271693284184b8175cab11d73a374ce4cce3202c91d68af612b62a3e","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"missing-dunder-method","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["pipeline.py"],"fixedSourceSha256":"cec14b7f9bc179d7b54c79b15ea3d4a0560661eca5329c23f0b33c834235cbd9","fixtureMetadataSha256":"0ce69947644a495a43e9ac6417184e761acbd00a1984ee6f15003e8dc6113159","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bm07-sklearn-pipeline-len"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"a1e504e6308f0d78a8da8269375cc64d155b448b9aea1b34ef8319a3abff0827","issueStatement":"# Bug: `Pipeline` does not implement `__len__`\n\nCalling `len(pipeline)` raises `TypeError: object of type 'Pipeline' has\nno len()` because the `__len__` method is not defined.\n\nAdd `__len__` to `Pipeline` so it returns `len(self.steps)`.\n","language":"python","ordinal":2,"primaryAggregate":true,"promptHashes":{"R":"1b5399be4da8449aa52dd8786c87e2a9e8f48547a1e825c1e3480cd3a6712a98","RN":"1b5399be4da8449aa52dd8786c87e2a9e8f48547a1e825c1e3480cd3a6712a98","U0":"cfcf855b33da4f9fa24260beca9bb9147b060eedc67f5a74615b3fe2735fefc7","U1":"c1ff9ed093f3062f78c5ccf68b6eda6821a5396f0a4fbbb52c80a4bb5182dcfd"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["pipeline.py"],"sourceSha256":"1e5bea10d9a183e94adc94da50daf517b427ec1e2944f1e7d55737ff53d4510e","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h03","title":"Cal high-affinity 03: Pipeline missing __len__ raises TypeError (GUIDED)","visibleEvaluatorSha256":"dbe4cdf7b856203e49b504d811d33437dd88f7e27737d5ac33e9f708e5ec968a","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"conditional-logic-wrong-operator","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["gate_check.cjs"],"fixedSourceSha256":"4fd700aa5b4c61777a7428f24afac0bbe28c9e7934d403ea0c476f238f0eba06","fixtureMetadataSha256":"7b7a30ea99d99cce7fc476e4e7d7f0def770ec9c50a4da0914fc5d6f6f8d1781","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-h05-v5-pair-19"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"3a05bc1e804e2e6ec0fcd99fbcb2c660f1845bca73f832c02a9916e2f14feb85","issueStatement":"# Bug: `canAccess` uses wrong logical operator\n\nThe `canAccess(user, resource)` function should only grant access when the\nuser is **both** an admin **and** active. An inactive admin or a non-admin\nactive user should not pass the gate (unless the resource is public).\n\nCurrently the condition uses `||` (OR) instead of `&&` (AND), allowing any\nactive user OR any admin to gain access.\n\nFix the logical operator in the `canAccess` condition so that access is\ngranted only when the user is simultaneously an admin and active.\n","language":"javascript","ordinal":3,"primaryAggregate":true,"promptHashes":{"R":"3cbd055e4962dff498355d26157b59b7a8709a1bdcf74de5d7176b584b5b7842","RN":"3cbd055e4962dff498355d26157b59b7a8709a1bdcf74de5d7176b584b5b7842","U0":"94851464b9381db1eabc6aaa0976f5aa2aa56cac64f5946638cdb036e2faa125","U1":"471c98cefd5d9fc118f9cb85930cc7b90435c51b022e1581f8e2f15e06f0b5b2"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["gate_check.cjs"],"sourceSha256":"cf4be787ef2cd483097e09062a7d2ce236904ccce26aacd7667fc9c4b32e3a2f","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h04","title":"Cal high-affinity 04: canAccess uses || instead of && (GUIDED)","visibleEvaluatorSha256":"639878ec6bd11a8f36418d097c9990c6cf3bff68a0d6103d5c1512a60053d62d","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"duplicate-results-from-join-filter","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["limit_choices.py"],"fixedSourceSha256":"12fd0d366d8f6e44fc706570f56e24570ce67321951f26237b030909bf15c2f7","fixtureMetadataSha256":"0c775a205338ea73f5d451b64f195ab3be3d164ae8b3501715125a20c8c84cd1","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bm10-django-limit-choices-duplicates"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"0c241d8e69a6de6830cc4df4015f617d4addccea43bc0b3f9d93098945c81167","issueStatement":"# Bug: `apply_limit_choices_to` produces duplicate records via `complex_filter`\n\nWhen `limit_choices_to` involves a join condition, `complex_filter` can\nreturn the same record multiple times. This causes duplicate options in\nform fields (e.g., a ModelChoiceField showing the same item twice).\n\nFix: replace `complex_filter` with a `filter()` call that uses an\nExists-like subquery, ensuring each record appears at most once.\n\nThe fixed version should:\n- Use `filter(exists_subquery=...)` instead of `complex_filter`\n- Produce no duplicates even when the criteria involve join-like conditions\n- Treat `None` and empty dict as no-op (don't filter)\n","language":"python","ordinal":4,"primaryAggregate":true,"promptHashes":{"R":"129abda1eb6f67680eb926b864ed4cd9fc89984712aa3b762c3cd57ceb71864c","RN":"129abda1eb6f67680eb926b864ed4cd9fc89984712aa3b762c3cd57ceb71864c","U0":"8f87442c2b7fcca27ada552312f8930cb83f4d94cc23d63556b2966c2ab4a8af","U1":"e1b69cd43347462c51d0e5c6cc8b43a954b29a610ccc6546282d9897263cdb9b"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["limit_choices.py"],"sourceSha256":"10b693dbb82350b9621c2a469e59a41c338ac70cb35bbd6cf2117ff08f401e0b","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h05","title":"Cal high-affinity 05: apply_limit_choices_to produces duplicates (GUIDED)","visibleEvaluatorSha256":"97d1e692fcf1e6e2cdd7bf9dd4647a6a90ccd9c7eafc8bff666bedae8125b619","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"missing-method-implementation","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["repr_printer.py"],"fixedSourceSha256":"7e6b07fb671629621aa3a407397221ee4bf3358b3a6313778b424c17ad03e46e","fixtureMetadataSha256":"a454dd9628e3443b987826a9aa5f9bca2862577227ca0fe3a5f863ce5487d0f0","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bm01-sympy-repr-dict-set"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"32e0ea3131ec3309f5dc6d50967e240504d48e3e00eeb9d7968af39a715161fd","issueStatement":"# Bug: `ReprPrinter` falls back to Python's `repr()` for dicts and sets\n\nThe `ReprPrinter` class dispatches to `_print_<TypeName>` methods but\nprinting a dict or set does not produce SymPy-style output; instead, the\nprinter falls back to Python's built-in `repr()`.\n\nFor example, `ReprPrinter().doprint({1: 2})` returns `'{1: 2}'` from\nPython's repr() rather than through the SymPy reprify machinery.\n","language":"python","ordinal":5,"primaryAggregate":true,"promptHashes":{"R":"661d50b45edbfc1948216f593122b97068b359bddcae4df28aa27c58943b849b","RN":"661d50b45edbfc1948216f593122b97068b359bddcae4df28aa27c58943b849b","U0":"6f5f1dacf14e1e146b7b381c3b5bc2fde602c7601c74321c846eda9b4281291f","U1":"92aebebc8f929aab6ed6044a69c9bbf96a6d1b93d877e3bfa2a72e85bf0e5023"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["repr_printer.py"],"sourceSha256":"e4a2541dc413286a6f099713273552b8c7479b68d4db8fa76d1c89f86344affc","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h06","title":"Cal high-affinity 06: ReprPrinter missing _print_dict and _print_set (REDUCED_GUIDANCE)","visibleEvaluatorSha256":"715bcb96310dd2a2ee73df0b36e4002137254f3b9fb88e880270380dc6f53a6b","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"off-by-one-boundary-condition","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["lognorm_clamp.py"],"fixedSourceSha256":"ed7ae5843f643b749687b984c35d46cdb4e81d8c704f833ccb2922bbb0a86ea4","fixtureMetadataSha256":"ccd4640d572be5888fdfcaa566784265aa57e97e39eaf6fb121502bf31c15e1e","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bm04-matplotlib-lognorm-boundary"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"244982bae1b9fee77627ce17d4eb282e52b4d6e20ef54ad7ea3c016ed5d53da9","issueStatement":"# Bug: `clamp_vmin_for_norm` raises `ValueError` when `s_vmin` is zero with `LogNorm`\n\nCalling `safe_log_normalize(LogNorm(), values, 0, 1)` raises a `ValueError`\ninstead of producing normalized output.\n\nThe `clamp_vmin_for_norm` function should protect against non-positive\n`s_vmin` values when using `LogNorm`, but the current boundary check does\nnot cover `s_vmin == 0`, so the zero case reaches the log computation\nand triggers an error.\n","language":"python","ordinal":6,"primaryAggregate":true,"promptHashes":{"R":"db79754223a3e276d02b7bf57b78519e5caf3e5cf68103ccd1d3b82f45df2bdf","RN":"db79754223a3e276d02b7bf57b78519e5caf3e5cf68103ccd1d3b82f45df2bdf","U0":"1b211bd28eddfa4ca8f04d9cd272e024bfc811802fcef6917819a5b269335892","U1":"38813ca317b58d6df506effa7dcefca646c04eb93332735a0fa2430d1c08335c"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["lognorm_clamp.py"],"sourceSha256":"2e96b95f761548b1c3f58ae7c991ef7aa0dd05443e0ed2a210a1a2e9af355e4b","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h07","title":"Cal high-affinity 07: LogNorm s_vmin=0 not clamped (REDUCED_GUIDANCE)","visibleEvaluatorSha256":"00b41d368f21a6d68e3186bd93f86f7207c30dfd25aa8d51625450219fb75f04","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"missing-parameter-propagation","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["engine_render.py"],"fixedSourceSha256":"b9b2d5fcd87c8ff61e12041c1835b44780de41b924571ef3affcba18031df7a2","fixtureMetadataSha256":"e5a65f791ec13d9b4ce1acba9a68dbc04fcd63833586ae4c5e03204f484e83b0","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bs01-django-engine-autoescape"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"4def0024b5ddc45483ce25bbffb1257fc5c8c67e0534ff7de2ef826fd4017a9b","issueStatement":"# Bug: `Engine.render_to_string` does not respect `autoescape=False`\n\nAn `Engine` created with `autoescape=False` still escapes HTML entities\nwhen `render_to_string` is called with a plain dict context.\n\nFor example:\n```python\nengine = Engine(autoescape=False)\nengine.add_template(\"t\", \"{{val}}\")\nresult = engine.render_to_string(\"t\", {\"val\": \"<b>hi</b>\"})\n# Expected: \"<b>hi</b>\"  — actual: \"&lt;b&gt;hi&lt;/b&gt;\"\n```\n","language":"python","ordinal":7,"primaryAggregate":true,"promptHashes":{"R":"fb577664d13bbf6d8af3bf0a47c3cf5626fbe3d4518fb3257a6a2f10267f3f3a","RN":"fb577664d13bbf6d8af3bf0a47c3cf5626fbe3d4518fb3257a6a2f10267f3f3a","U0":"c4213f4ae40e511895729ade5f3bf88e917f6a2b41b2b5a0eb7eec3b8d7585e2","U1":"7c3584b45f709853f4bb631e8ba5c7eccd9728161576b7514579d674e1b23df0"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["engine_render.py"],"sourceSha256":"91d92eee181925800021f8408582477d8279a90a4b43e983f66a3700ae33fcb5","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h08","title":"Cal high-affinity 08: Engine.render_to_string ignores autoescape (REDUCED_GUIDANCE)","visibleEvaluatorSha256":"92d0f684ea8ee9ed4d0aca7bce9008b66f595c85d425ad1f19c27d885e67f3d5","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"strict-normalize-wrong-case-direction","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["canonicalize.cjs"],"fixedSourceSha256":"49280ee561ff14e56518c20a146501d812f2f332c009914ab8f9cc11313cbcc7","fixtureMetadataSha256":"54d03e876b55b8f8f791c233ccbe815d905e9a9f165d2bd1b849378046702423","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-h09-v5-pair-20"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"11009006baa9f02fa947e8df058b4043c9ea583d1e4e6967edef030ae1832308","issueStatement":"# Bug: `canonicalize` strict mode produces wrong-case output\n\nWhen `mode=\"strict\"`, `canonicalize(input)` should produce a lowercase\nunderscore-separated canonical form. For example,\n`canonicalize(\"Ada Lovelace\", \"strict\")` should return `\"ada_lovelace\"`.\n\nHowever, the current output is `\"ADA_LOVELACE\"` — the wrong case.\n","language":"javascript","ordinal":8,"primaryAggregate":true,"promptHashes":{"R":"8c695516f365ae6434a04b3430dd671ca858aca96e958eb678f2021c6659b8ef","RN":"8c695516f365ae6434a04b3430dd671ca858aca96e958eb678f2021c6659b8ef","U0":"d3d79535240160544429ab2df32c2a3333fb36d709671beac812d0e03f68c63b","U1":"5caf0a5cc0e8eba94810fdb094221c32112730e801fb5772c31be5a5f2405533"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["canonicalize.cjs"],"sourceSha256":"72de8a13ab779cbc9ac50f6693ec9965c59d2ea2ebc5c63e49a5e303bf7f7f12","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h09","title":"Cal high-affinity 09: canonicalize strict mode wrong case direction (REDUCED_GUIDANCE)","visibleEvaluatorSha256":"c63512cfcd4bbb74904c14fb56affa85601c14894f290ea240359794d9b7f64f","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"HIGH_REPLACEMENT_AFFINITY","bugClass":"off-by-one-boundary-condition","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["floatformat.py"],"fixedSourceSha256":"d92620f4420dc4bca96baa09a2edc55c20aced62e0904648f862f8182b1659b2","fixtureMetadataSha256":"32406bbd1e882a904eddec1e5e87f3bcea997e6070feff7c54de854b6e38e963","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","selectionAlgorithm":"xorshift-shuffle-lex-sorted-primary-taskIds-take-10","selectionSeed":"qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576","upstreamProtocol":"qwen3-32b-v7-paired-interface-expansion-v2","upstreamTaskId":"exp-bs03-django-floatformat-zero"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"36737ca27203328f2fce25f5d8cc362796d55bb9780668bf145d24607225c17d","issueStatement":"# Bug: `floatformat` raises `ValueError` for `\"0.00\"` with arg `0`\n\nCalling `floatformat('0.00', 0)` raises a `ValueError` (precision must be\npositive) instead of returning `\"0\"`.\n\nThe function should handle the case where the input has no significant\ndecimal digits and the precision argument is zero, but currently it\nmisclassifies this case and attempts to quantize with zero places in a\ncontext that raises an error.\n","language":"python","ordinal":9,"primaryAggregate":true,"promptHashes":{"R":"8e303679ea91567c3e6f6718fda76b2ef1f208765696427ca01fc6de1283cc16","RN":"8e303679ea91567c3e6f6718fda76b2ef1f208765696427ca01fc6de1283cc16","U0":"90312de32fe4a98933b96f00a6a0637302372f0787f21d8c7c64d36d30e5c587","U1":"2495519e839d9da1af0414391e318d1921341e3b08a30ef109a381a86ccbcfba"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["floatformat.py"],"sourceSha256":"b56985c7888f93f924c6efe7fede43fc51c85fd3d12a6981ad956a7bf6977b9e","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-h10","title":"Cal high-affinity 10: floatformat crashes on 0.00 with arg 0 (REDUCED_GUIDANCE)","visibleEvaluatorSha256":"6c287588ea53fae1274008ea103cc985e56d79844a2b702d6c3c7b2a2c5fb37e","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"multi-hunk-wrong-operator-and-case","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["normalize.cjs"],"fixedSourceSha256":"7ccf3a7631d6cbf49dcd9759a64274d4fe20a960cfe57ac8dd3e7050927357b1","fixtureMetadataSha256":"3c6e9b2425c8586ccd3da7ffb8896829d8c56adcf77126196a7f45488c67d5b6","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"multi-hunk-wrong-operator-and-case"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"24042fa3454a31c31d9e9fda181d6353331d13ae11362073bc4634f6040aad0e","issueStatement":"# Bug: `normalize` compact mode has two errors\n\nThe `normalize(input, \"compact\")` function should tokenize the input,\njoin tokens with a single space, and return the result in lowercase.\n\nThere are two bugs in the compact mode branch:\n1. Tokens are joined with `\",\"` (comma) instead of `\" \"` (space).\n2. The result is transformed to uppercase instead of lowercase.\n\nFor example, `normalize(\"Hello  World\", \"compact\")` should return\n`\"hello world\"` but currently returns `\"HELLO,WORLD\"`.\n","language":"javascript","ordinal":10,"primaryAggregate":true,"promptHashes":{"R":"f2373fcdae853ca48b90fe345ed9363d076aa93f2b29cd102493f4e40f9b6dfc","RN":"f2373fcdae853ca48b90fe345ed9363d076aa93f2b29cd102493f4e40f9b6dfc","U0":"e41c06bfda6e305bada603a458567a3d35143a9337aa3414fa026f444344ab30","U1":"98cab0e127808e1bcaa0a40dbd770ecb0cbf579b99f3ce1c759ccd36a96bdfc6"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["normalize.cjs"],"sourceSha256":"a48c789030387a76ed05ae477a062af4e869b510a44bce54cabded687b5dcd67","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l01","title":"Cal low-affinity 01: normalize compact mode has two bugs (multi-hunk, GUIDED)","visibleEvaluatorSha256":"f6887ccab03a6bef42e2168585aad927ec15903848b46a6f85ff94183c123954","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"multi-hunk-wrong-aggregation","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["stats.py"],"fixedSourceSha256":"a93b3ea5cb79d83a48aa66d5b30208f344621abdb75844bb0bc9552a38bacd6d","fixtureMetadataSha256":"26afc920ae4837f85d5a5974d3340014d31ee81436a974551d965bd270eec5bd","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"multi-hunk-wrong-aggregation"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"9208724843440b4664946c815718f7fabca98afd0af27e40db3d9ecbddb81090","issueStatement":"# Bug: `summary` has two errors computing low and mean\n\nThe `summary(items)` function should return the minimum value as `\"low\"`,\nthe maximum as `\"high\"`, and the arithmetic mean as `\"mean\"`.\n\nTwo bugs exist:\n1. `low` is computed as `max(items)` instead of `min(items)`.\n2. `mean` is computed by dividing by `2` instead of `len(items)`.\n\nFor example, `summary([1, 2, 3, 4])` should return\n`{\"low\": 1, \"high\": 4, \"mean\": 2.5}` but currently returns\n`{\"low\": 4, \"high\": 4, \"mean\": 5.0}`.\n","language":"python","ordinal":11,"primaryAggregate":true,"promptHashes":{"R":"6427c9a74e5a0d3ccb9672e27a035ec6e8f60b680c3b003e0f106fd16dbed21e","RN":"6427c9a74e5a0d3ccb9672e27a035ec6e8f60b680c3b003e0f106fd16dbed21e","U0":"464615acd70ca060fa345753dddd75a7b9061c5e4fb3e053117f09f5b566877b","U1":"ea2815cb1342de65b9e5d22472a23801c1b2c3102e5cb27b1b78d6eccb159842"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["stats.py"],"sourceSha256":"5c3bcfa4b2ef1b35c0b82297a4e529c19165478c49d66182dfb30dff463505e1","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l02","title":"Cal low-affinity 02: summary uses max for low and wrong divisor for mean (multi-hunk, GUIDED)","visibleEvaluatorSha256":"d3062bbe28586a1d404a132687a59b983a625067403516f45b64d3cbe6a2fc13","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"missing-null-guard-insertion","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["parser.cjs"],"fixedSourceSha256":"685fd1bfcf02da57d8f5f7eb3b5dc617439412607dc388a1e357b07756e26785","fixtureMetadataSha256":"610d285779da8f87fc9d71ddd030be0f6a2dc71fc7fd8517c87d1b3d8b1ff0db","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"missing-null-guard-insertion"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"7fa2054fbd96a3e1f93a28b7a101874a399a28cf85df3238c4a395e879eb28ed","issueStatement":"# Bug: `parseConfig` crashes with `TypeError` when given `null` or `undefined`\n\nCalling `parseConfig(null)` throws `TypeError: Cannot read properties of\nnull (reading 'split')` because the function calls `input.split(\"\\n\")`\nwithout first checking whether the input is null or undefined.\n\nAdd an early return at the beginning of `parseConfig` that returns an\nempty object `{}` when `input` is `null` or `undefined`.\n","language":"javascript","ordinal":12,"primaryAggregate":true,"promptHashes":{"R":"45ab3ac44ccea0c335bd06c3181d964a0ecfdaa671473e9f2179c86630fe032a","RN":"45ab3ac44ccea0c335bd06c3181d964a0ecfdaa671473e9f2179c86630fe032a","U0":"7cad0c0e36f82ea2efb627ed654fee8c6e5407ac79ca7bd160acc3bd5a8f10d0","U1":"0b9eaba3b20fed321d511077e053f2bb5c0e700c503f50afdfdea5b06d070232"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["parser.cjs"],"sourceSha256":"ca4c7c514bf825b1e5f67b3700ec5ceb2ca0210731c4a6789303f0045e0246b9","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l03","title":"Cal low-affinity 03: parseConfig crashes on null/undefined input (insertion, GUIDED)","visibleEvaluatorSha256":"fc2f2fa1f50b0cd4cc0d5cae5d67203ca653e84e571ddb58311254c8fddb95a9","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"missing-range-validation-insertion","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["validator.py"],"fixedSourceSha256":"3dc8eda813e01145d9bd7d1f7a539938fe92bcc6d28d646a69408cdae7d833f6","fixtureMetadataSha256":"f954c1aa54385a952735b8c97cdb653e697d77ab81850b29c26a3e4f1124233b","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"missing-range-validation-insertion"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"3aa409be7844e7a5bf78de20724cbef3fd5f6255a98a54082a09d2c8ffcc7dd9","issueStatement":"# Bug: `validate_config` does not check port range\n\n`validate_config` verifies that `port` is an integer but does not check\nthat it falls within the valid TCP port range (1 to 65535 inclusive).\n\nAs a result, ports like `0`, `-1`, or `99999` pass validation even though\nthey are invalid.\n\nAdd a range check after the type check: if `port < 1` or `port > 65535`,\nreturn `(False, \"port must be in range 1..65535\")`.\n","language":"python","ordinal":13,"primaryAggregate":true,"promptHashes":{"R":"66e74c3f17077d788746b92acc5174a071c015785917cd08c37f0b733a8808a7","RN":"66e74c3f17077d788746b92acc5174a071c015785917cd08c37f0b733a8808a7","U0":"fe0ad32559248ad10fd02f1ec42127c2aad41c7233711137e0e5a631cde72407","U1":"9e9cbcbfd9b96468de26ac6ab4651e9fd21cf661b9c5612c5389ad1d67014439"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["validator.py"],"sourceSha256":"1596e86b59b08faacc0fa7ae026f12ac78a2bde741f0071c6573eaf3616bbee1","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l04","title":"Cal low-affinity 04: validate_config does not check port range (insertion, GUIDED)","visibleEvaluatorSha256":"adcd074a4c8bb7deb7b6b3159a0a7b9ab37bb0655e3d5af3b37c5d2f0cbf6bd9","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"spurious-early-return-deletion","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["cleanup.cjs"],"fixedSourceSha256":"57ded9e51d5efa33c4f00429553f3cabc262cacd8f4d8590324c102f58520b79","fixtureMetadataSha256":"389ca32c76ccbaadb5883ad71ea9d75dd47e1ac383495b07575f44f97663867c","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"spurious-early-return-deletion"},"guidanceClass":"GUIDED","hiddenEvaluatorSha256":"c52b751beb61c584e90ae1b01990b0861cf23cf8d85c6521d0f7e1e8184bfbd8","issueStatement":"# Bug: `removePattern` always returns the original text unchanged\n\n`removePattern(text, pattern)` is supposed to remove all occurrences of\n`pattern` from `text`, but it returns `text` unmodified every time.\n\nThe function contains a spurious `return text;` statement before the\ncorrect `return text.replace(re, \"\");` line, causing the replacement to\nnever execute.\n\nRemove the spurious `return text;` line so that the replacement executes.\n","language":"javascript","ordinal":14,"primaryAggregate":true,"promptHashes":{"R":"f33f629ccd040b8ac7c4fa3a2d5f9fe2f18646940dbf03687883e23a7786b442","RN":"f33f629ccd040b8ac7c4fa3a2d5f9fe2f18646940dbf03687883e23a7786b442","U0":"ac90b529bc4aee43bcc7e6e1a8119f414449eb48d28332bf98b765520c279b2b","U1":"465e45ba4013bc3f0f21f79266290895cf1ac8157e3c2cef61ba724ca4206933"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["cleanup.cjs"],"sourceSha256":"bd66ae73389284de6d6caeda78b1b4a9f400c879c5efdee959a8bd83db1a7254","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l05","title":"Cal low-affinity 05: removePattern always returns unchanged text (deletion, GUIDED)","visibleEvaluatorSha256":"4627d0ff8d2aa4b4493aa9438d7336e161ec94e8c6eb1e913a6886bd7996baf7","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"spurious-extra-condition-deletion","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["filter_data.py"],"fixedSourceSha256":"3fda556470fd3ffddc3f9c6facaf7f4a736583fb6bebecb7f51a85d3e4bf24f4","fixtureMetadataSha256":"ed5f065eaaebb1ba9002e662abc58a13dc916bf5886a8dec43f207f67d3c20bc","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"spurious-extra-condition-deletion"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"b9e83eee77f97a8588c150651712a05957d9af0782b4c54a6b11a8441c44528a","issueStatement":"# Bug: `filter_active` excludes valid active records that have no `tag` field\n\n`filter_active(records)` should return all records where `active` is `True`,\nregardless of whether any other fields are present.\n\nHowever, some active records are incorrectly excluded: active records that\ndo not have a `tag` field are silently dropped.\n\nFor example:\n```python\nrecords = [{\"active\": True, \"name\": \"Alice\"}]\nfilter_active(records)  # Returns [] instead of the expected [{\"active\": True, \"name\": \"Alice\"}]\n```\n","language":"python","ordinal":15,"primaryAggregate":true,"promptHashes":{"R":"e49f62fa7f180ae66b806ea5a55d245ca31593a944e1e6795e48c8d1c7e5e3e7","RN":"e49f62fa7f180ae66b806ea5a55d245ca31593a944e1e6795e48c8d1c7e5e3e7","U0":"d7879ade4a967498eb19754d1b805c047b48611e9cc9736ad9a7080d1ffea397","U1":"0cc8811ce8648eea6659bd7d626add87f65b0ef6bd2ffb08705a98ab88f1e043"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["filter_data.py"],"sourceSha256":"3429f3d6a22ec003880f06161cfdf4b86a140521510bbbce818b7cb765ed956c","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l06","title":"Cal low-affinity 06: filter_active excludes active records without tag (deletion, REDUCED_GUIDANCE)","visibleEvaluatorSha256":"09015c9823841d1cf3090b829f5d485377fb4eb9885b136e6492e2c8b6458272","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"wrong-return-value-repeated-text","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["tokenizer.cjs"],"fixedSourceSha256":"ca38bb7babe8c381c54f1d0c460fadaa2e3ca2ecb04f181db255f0560b2b232d","fixtureMetadataSha256":"6887f29a2bad4e2a77932bf0d7c528baf49a51b4d43397e3b6d0dc0950fd478d","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"wrong-return-value-repeated-text"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"0f70143ce6978bc6669031e982377c11e3164cda0e98b88ef36645951df4ff06","issueStatement":"# Bug: `tokenize` returns `null` for empty strings instead of `[]`\n\n`tokenize(text)` should return an empty array `[]` when given an empty or\nwhitespace-only string, but currently returns `null`.\n\nThis causes `countWords(\"\")` to return `null` instead of `0`, which breaks\ncallers that expect an array for any valid string input.\n\nNote: `null` is the correct return value for `null` input or non-string input\n(invalid inputs). Empty strings are valid inputs and should return `[]`.\n","language":"javascript","ordinal":16,"primaryAggregate":true,"promptHashes":{"R":"4b75bfc4cd1b6e2eee9c365cf36601c51839a825c84ea329de5abb927000f816","RN":"4b75bfc4cd1b6e2eee9c365cf36601c51839a825c84ea329de5abb927000f816","U0":"9d5019cbf36e1ae082dbbf237f5c1f3f131dd65254bb9241dfd8088fafe558fc","U1":"a4cf31b6e2dbf22ba42781846c793fe8b1ce9529858e1cad065ae2c88bcb7f59"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["tokenizer.cjs"],"sourceSha256":"5503fde9c8eb254481043818186167b01053d49d81ab134e33d63193a85d8618","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l07","title":"Cal low-affinity 07: tokenize returns null for empty strings (repeated-text ambiguity, REDUCED_GUIDANCE)","visibleEvaluatorSha256":"46a44b11d02921ca4ff74d51c1ad5024b7dd758ef0baec7eb95d8528f2ab975c","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"wrong-return-value-repeated-text","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["encoder.py"],"fixedSourceSha256":"5eb5bbefa7136fd03c3d93bd024cd8e0fde870fd8e6e075be5fe7321d7065da0","fixtureMetadataSha256":"a1621bdac533a3ab13e3f5d14e55df3a34b4ea1f1061d47f8c2e653d89150f7c","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"wrong-return-value-repeated-text"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"6479d8bab5fe824db628866a41e38a0f4864791e54348e6941088a2ca86b6387","issueStatement":"# Bug: `encode` returns empty string for single-character input\n\n`encode(text)` should produce a run-length encoded string where each run\nis represented as `count + character`. A single character `\"a\"` should\nencode as `\"1a\"`, but currently returns `\"\"`.\n\nThe other early-return paths (`not isinstance(text, str)` → `\"\"`, and\n`len(text) == 0` → `\"\"`) are correct. Only the single-character branch\nis wrong.\n","language":"python","ordinal":17,"primaryAggregate":true,"promptHashes":{"R":"54e9426f1b808f77c674654695972ece129265f54a8b49001a8c14158d3134d4","RN":"54e9426f1b808f77c674654695972ece129265f54a8b49001a8c14158d3134d4","U0":"3dfbe7505bbe9057c923f0660896e5fe93d2e81011b0743517762fc1f8933cff","U1":"46d56500f29b52bf4286dda048fe5478d282f619e0a49623034ff9a0c054f960"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["encoder.py"],"sourceSha256":"65db2333d36743d2f7fc236143c58bab05617f31cc097c5a8423103186d115b1","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l08","title":"Cal low-affinity 08: encode returns empty for single-char input (repeated-text ambiguity, REDUCED_GUIDANCE)","visibleEvaluatorSha256":"71f4e7866e12e68046198aba575ec4b740421c6effd900816c0f673bcf8625d7","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"wrong-constant-value-distant-usage","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_AND_R","filesAllowlist":["config.cjs"],"fixedSourceSha256":"823a827c9854efa78de45b75e30327a9adeed3441f6addbc4cd33aef3930a860","fixtureMetadataSha256":"6fee18f405e7a5daa369e88d9a48d79c4be47b606013a477cb00be09d42f0e4f","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"wrong-constant-value-distant-usage"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"9c8b4847195dc3beec156d850a7f4a267d00a650c04d01704118c93f8e59be2e","issueStatement":"# Bug: `DEFAULT_TIMEOUT_MS` is set to `100` instead of `5000`\n\nThe default timeout should be 5000 milliseconds (5 seconds), but\n`DEFAULT_TIMEOUT_MS` is set to `100`. This constant is used in two places:\nas the default value in `getDefaultConfig()` and as the minimum threshold\nin `createConfig()`. Both references produce incorrect behavior with the\nwrong value.\n\nAfter the fix, `getDefaultConfig().timeout` should be `5000`, and\n`createConfig()` should also use `5000` as the minimum acceptable timeout.\n","language":"javascript","ordinal":18,"primaryAggregate":true,"promptHashes":{"R":"4dc0142a20577ed7bdff5a2f6cbc923a2a51d60cc32e61ab36b3b908b628efa6","RN":"4dc0142a20577ed7bdff5a2f6cbc923a2a51d60cc32e61ab36b3b908b628efa6","U0":"dd0a84d83bfa631d8b980697ffaff46a8fe0a18afbaea88f810051ecb59ac081","U1":"4c1935e7f7c16ee140202ec485369d938a8e38440c655498d5644f62222556f5"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["config.cjs"],"sourceSha256":"8cb9160ebea0a834e3437e434fb036d3f26d4172c669a365978a2b7e8ed75fcf","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l09","title":"Cal low-affinity 09: DEFAULT_TIMEOUT_MS wrong value affects two usages (coordinated distant, REDUCED_GUIDANCE)","visibleEvaluatorSha256":"379250f59661e3ffd5818554701cb1b193cd0b20213e892e3f2a82a1562ba27e","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}
+{"affinityStratum":"LOW_REPLACEMENT_AFFINITY","bugClass":"wrong-constant-value-cross-file","contentReleaseDecision":"FULL_CONTENT_EXPORT_ALLOWED","expressibilityClass":"EXPRESSIBLE_BY_U_ONLY","filesAllowlist":["registry.cjs","registry-defaults.cjs"],"fixedSourceSha256":"3eb29d648c14f18134968b17a85fa5931c5d28cf0c6d9336f49a4635d0a45df1","fixtureMetadataSha256":"37e68709d5a35610c11a30915f5b3658c283bd58932a6951af8753c32a7e5f06","fixtureProvenance":{"calibrationStudy":"qwen3-32b-v7-four-arm-interface-calibration-v1","source":"hand-crafted","stratumConstraints":"wrong-constant-value-cross-file"},"guidanceClass":"REDUCED_GUIDANCE","hiddenEvaluatorSha256":"5b4384c412cf62e8fd8240f46b5e905ebab7d8007b9eb6268483a7736a9f70ae","issueStatement":"# Bug: `DEFAULT_VALIDATION_MODE` is `\"permissive\"` instead of `\"strict\"`\n\nThe plugin registry should default to strict validation mode, which enforces\nthat registered plugins have a non-empty string name and an `execute()`\nmethod. However, `DEFAULT_VALIDATION_MODE` in `registry-defaults.cjs` is\nset to `\"permissive\"`, so invalid plugins can be registered silently.\n\nThe fix requires changing `DEFAULT_VALIDATION_MODE` from `\"permissive\"` to\n`\"strict\"` in `registry-defaults.cjs`. Because `registry.cjs` re-exports\n`DEFAULT_VALIDATION_MODE`, that file's export will also reflect the correct\nvalue after the fix.\n\nAfter the fix, `createRegistry().getValidationMode()` returns `\"strict\"`,\nand attempting to register a plugin without an `execute()` method throws a\n`TypeError`.\n","language":"javascript","ordinal":19,"primaryAggregate":false,"promptHashes":{"R":"fa210735f600fcc7e9424ad1ea6603134fba02e46091e337474c2b869748ebf1","RN":"fa210735f600fcc7e9424ad1ea6603134fba02e46091e337474c2b869748ebf1","U0":"37f04579da44e0db4e2830c64a24a5141664bb26d30229c4b3ecc4511ec34efd","U1":"06212d43a9fb77e42125efeef3692eb18acbfd5fa4f125d91eadc0c963612366"},"protocolId":"qwen3-32b-v7-four-arm-interface-calibration-v1","rPromptEqualsRnPrompt":true,"recordKind":"PUBLIC_CALIBRATION_TASK_RECORD","schemaVersion":1,"sourcePaths":["registry.cjs","registry-defaults.cjs"],"sourceSha256":"3eb29d648c14f18134968b17a85fa5931c5d28cf0c6d9336f49a4635d0a45df1","studyId":"workledger-structured-edit-interface-study-v0.1","taskId":"cal-l10","title":"Cal low-affinity 10: DEFAULT_VALIDATION_MODE wrong across two files (two-file, REDUCED_GUIDANCE)","visibleEvaluatorSha256":"018c4ba27a4cda7abdfebacee8ac927d277806b39defbc577557c606bcb981b3","withheldFields":["brokenSource","fixedSource","visibleEvaluatorSource","hiddenEvaluatorSource","privateFixturePath"]}

docs/LIMITATIONS.md

@@ -0,0 +1,3 @@
+# Limitations
+
+This alias mirrors docs/limitations.md. The release package is a bounded offline controlled interface study with one attempt per cell, route/inference confounds, no production-readiness claim, and no upstream-source relicensing claim.

docs/METHODOLOGY.md

@@ -0,0 +1,3 @@
+# Methodology
+
+This alias mirrors docs/methods.md for consumers that expect an uppercase methodology path. The release root is an offline controlled-evaluation package built from sanitized records only, with no provider transport, upload, DOI, tag, or external publication action.

docs/REPRODUCTION.md

@@ -0,0 +1,3 @@
+# Reproduction
+
+Rebuild the package with the repository export builder and compare `provenance/MANIFEST.json` hashes. The public package supports offline inspection of sanitized records; it is not a live provider execution harness and performs zero provider requests during rebuild.

docs/SECURITY_AND_PRIVACY.md

@@ -0,0 +1,3 @@
+# Security and privacy
+
+The public export excludes hidden evaluator source, raw provider response bodies, authorization material, idempotency keys, request bodies, private filesystem paths, SQLite ledgers, and reasoning traces. Leakage checks are recorded in `provenance/PUBLIC_EXPORT_LEAKAGE_CHECK.json`.

docs/limitations.md

@@ -0,0 +1,10 @@
+# Limitations
+
+- One attempt was recorded per cell.
+- Provider route and inference settings are confounded with model identity.
+- Qwen3-32B is an external anchor, not a jointly randomized member of the paid multimodel denominator.
+- Qwen3 8B has paired coverage 18/19 and no registered interface-advantage replication.
+- DeepSeek V4 Pro is infrastructure-unassessable because 38/38 cells failed at the privacy/data-policy route layer with empty final answers; do not treat it as a coding-performance denominator.
+- The model-strength interaction is descriptive only, and CROSS_MODEL_GENERALIZATION_SIGNAL is false.
+- The package is not a production-readiness claim and not a live controlled-evaluation runner.
+- Low-affinity tasks are newly authored for calibration where supported; high-affinity tasks are adapted or reconstructed from upstream issue/task patterns with upstream bytes withheld.

docs/methods.md

@@ -0,0 +1,7 @@
+# Methods
+
+The release dataset is copied from the sanitized public package and augmented only with release-candidate metadata, cards, audit files, and manifests. It is provider-free and does not construct network transport.
+
+For multimodel records, strict native and one-outer-fence-normalized policies are reported on identical response bytes. DeepSeek V4 Pro cells are infrastructure-unassessable privacy/data-policy route failures with empty final answers, not coding-performance failures.
+
+Low-affinity calibration tasks are newly authored for calibration where supported by provenance. High-affinity calibration tasks are adapted or reconstructed from upstream issue/task patterns; upstream source bytes and evaluator bytes are withheld.

docs/suggested-tag.md

@@ -0,0 +1,5 @@
+# Suggested tag
+
+`study/structured-edit-interface-v0.1`
+
+This file documents release preparation only. The build and verification scripts do not create tags or releases.

provenance/BUILD_PROVENANCE.json

@@ -0,0 +1,21 @@
+{
+  "generatedBy": "scripts/build-workledger-public-study-export.ts",
+  "offlineOnly": true,
+  "offlineReconciliationLabel": "OFFLINE_DERIVED_RECONCILIATION_NOT_A_REPLACEMENT_FOR_REGISTERED_ENDPOINT",
+  "providerRequests": 0,
+  "recordKind": "PUBLIC_EXPORT_BUILD_PROVENANCE",
+  "schemaVersion": 1,
+  "sourceArtifactLabels": [
+    "CALIBRATION_TASK_BANK_MANIFEST.json",
+    "AGGREGATE_ANALYSIS.json",
+    "PR48_CELL_STAGE_REPLAY.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/PER_MODEL_ANALYSIS.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/analysis/AGGREGATE_ANALYSIS.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/DECISION_CLASSIFICATION.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/cells/*/attempt-1/raw-response.txt"
+  ],
+  "sourceEvidenceTip": "c8e876133ed2745f23bc14d00aca79d68e83d590",
+  "sourceProtocolId": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+  "studyId": "workledger-structured-edit-interface-study-v0.1",
+  "transportConstructed": false
+}

provenance/LICENSE_AUDIT.json

@@ -0,0 +1,294 @@
+{
+  "decisions": [
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "material": "calibration task metadata and issue statements",
+      "provenanceFieldsRetained": [
+        "calibrationStudy",
+        "selectionAlgorithm",
+        "selectionSeed",
+        "upstreamProtocol",
+        "upstreamTaskId"
+      ],
+      "rationale": "Low-affinity calibration task wording is newly authored for calibration where supported by provenance; high-affinity wording is adapted/reconstructed from upstream issue/task patterns. Upstream source bytes, visible evaluator bytes, hidden evaluator bytes, and private fixture paths are withheld.",
+      "taskCount": 20
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "material": "PR48 calibration and multimodel live final assistant answers",
+      "rationale": "Final assistant answers are the public study observations; private provider bodies, route details, identifiers, and reasoning are withheld."
+    },
+    {
+      "decision": "METADATA_AND_HASHES_ONLY",
+      "material": "expansion and pilot task banks",
+      "rationale": "Prior studies are summarized only as claim metadata because licenses are not asserted here."
+    },
+    {
+      "decision": "METADATA_AND_HASHES_ONLY",
+      "material": "visible evaluator source",
+      "rationale": "Evaluator hashes are sufficient for this public export; evaluator source is not required for the replay-only package."
+    },
+    {
+      "decision": "WITHHOLD",
+      "material": "hidden evaluator source",
+      "rationale": "Hidden evaluator source is never included in the public export."
+    },
+    {
+      "decision": "WITHHOLD",
+      "material": "provider transport evidence",
+      "rationale": "Request envelopes, response identifiers, raw provider bodies, credentials, and reasoning are private execution evidence."
+    },
+    {
+      "decision": "WITHHOLD",
+      "material": "SQLite ledgers and private artifact paths",
+      "rationale": "Public provenance uses hashes and package-relative paths only."
+    }
+  ],
+  "licenseAssertions": {
+    "huggingFaceDatasetCardAndPublicReleaseMetadata": "cc-by-4.0",
+    "huggingFaceSpaceCodeAndTooling": "mit",
+    "publicDatasetMetadataAndRecords": "cc-by-4.0",
+    "upstreamSourceAndEvaluatorBytes": "not relicensed; withheld where upstream/license treatment is not public-release material"
+  },
+  "perTaskFixtureProvenance": [
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bs06-django-ordered-set-reversed"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h01"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bm09-pytest-mock-sentinel"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h02"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bm07-sklearn-pipeline-len"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h03"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-h05-v5-pair-19"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h04"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bm10-django-limit-choices-duplicates"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h05"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bm01-sympy-repr-dict-set"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h06"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bm04-matplotlib-lognorm-boundary"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h07"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bs01-django-engine-autoescape"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h08"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-h09-v5-pair-20"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h09"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "selectionAlgorithm": "xorshift-shuffle-lex-sorted-primary-taskIds-take-10",
+        "selectionSeed": "qwen3-32b-v7-four-arm-interface-calibration-v1|high-affinity|807ecca064acae40f754532b3334bda9b917bb7d22410144b929ca6dcb595576",
+        "upstreamProtocol": "qwen3-32b-v7-paired-interface-expansion-v2",
+        "upstreamTaskId": "exp-bs03-django-floatformat-zero"
+      },
+      "releaseProvenanceWording": "adapted/reconstructed from upstream issue/task patterns; upstream source/evaluator bytes withheld",
+      "taskId": "cal-h10"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "multi-hunk-wrong-operator-and-case"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l01"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "multi-hunk-wrong-aggregation"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l02"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "missing-null-guard-insertion"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l03"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "missing-range-validation-insertion"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l04"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "spurious-early-return-deletion"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l05"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "spurious-extra-condition-deletion"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l06"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "wrong-return-value-repeated-text"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l07"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "wrong-return-value-repeated-text"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l08"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "wrong-constant-value-distant-usage"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l09"
+    },
+    {
+      "decision": "FULL_CONTENT_EXPORT_ALLOWED",
+      "provenance": {
+        "calibrationStudy": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+        "source": "hand-crafted",
+        "stratumConstraints": "wrong-constant-value-cross-file"
+      },
+      "releaseProvenanceWording": "newly authored for calibration",
+      "taskId": "cal-l10"
+    }
+  ],
+  "recordKind": "PUBLIC_EXPORT_LICENSE_AUDIT",
+  "releaseLicensePolicy": {
+    "datasetLicenseIdentifier": "cc-by-4.0",
+    "noUpstreamRelicensing": true,
+    "privateEvidenceWithheld": true,
+    "spaceCodeLicenseIdentifier": "mit"
+  },
+  "schemaVersion": 1,
+  "studyId": "workledger-structured-edit-interface-study-v0.1",
+  "summary": {
+    "FULL_CONTENT_EXPORT_ALLOWED": 1,
+    "METADATA_AND_HASHES_ONLY": 2,
+    "WITHHOLD": 3
+  },
+  "taskWordingPolicy": {
+    "highAffinity": "High-affinity calibration tasks are adapted or reconstructed from upstream issue/task patterns; upstream source bytes and evaluator bytes are withheld.",
+    "licenseTreatment": "Public dataset metadata and public release records are released under CC-BY-4.0; Space code/tooling is MIT. Upstream source bytes, evaluator bytes, private evidence, and withheld fixtures are not relicensed by this release.",
+    "lowAffinity": "Low-affinity calibration tasks are newly authored for this calibration when their release provenance records show the hand-crafted source policy."
+  }
+}

provenance/MANIFEST.json

@@ -0,0 +1,163 @@
+{
+  "entries": [
+    {
+      "bytes": 15058,
+      "path": "data/aggregate_results.jsonl",
+      "sha256": "81bf95b543e8d474c14c7b350978fce7bf541f0940c7720ee2ab89c40ff63c87"
+    },
+    {
+      "bytes": 20118,
+      "path": "data/aggregate-results.json",
+      "sha256": "6c8daea0e22c7aad2a93a37a533ebc571193ab01301e4eda329293469d57ca38"
+    },
+    {
+      "bytes": 10346,
+      "path": "data/claim_survival.jsonl",
+      "sha256": "3341108d1e51f0b1a249e7e7fdc9edce7fef59598e4817b4a1364964236a49c9"
+    },
+    {
+      "bytes": 11765,
+      "path": "data/claim-survival.json",
+      "sha256": "45aa24c6a3c0676fe888e6aa49893f4c05b42455e115342aa64cc3b738fa0c94"
+    },
+    {
+      "bytes": 667543,
+      "path": "data/generations.jsonl",
+      "sha256": "78363d7c143d289b092dde2c1ae1c674529d621573ff5c53d4c5618984b00299"
+    },
+    {
+      "bytes": 3177,
+      "path": "data/model-metadata.json",
+      "sha256": "16acadc395b0f211b65bae3ad89ea94675ec6645b16e73a6d6de11ad1bd8b628"
+    },
+    {
+      "bytes": 667543,
+      "path": "data/stage-outcomes.jsonl",
+      "sha256": "78363d7c143d289b092dde2c1ae1c674529d621573ff5c53d4c5618984b00299"
+    },
+    {
+      "bytes": 1165,
+      "path": "data/studies.jsonl",
+      "sha256": "80a2c92bc7657c91405f1fc6d89a33fe2b5e4547b22539fa83baadee90cbdc6d"
+    },
+    {
+      "bytes": 1165,
+      "path": "data/study_metadata.jsonl",
+      "sha256": "80a2c92bc7657c91405f1fc6d89a33fe2b5e4547b22539fa83baadee90cbdc6d"
+    },
+    {
+      "bytes": 46642,
+      "path": "data/tasks.jsonl",
+      "sha256": "416b1602fd0aedd2c34274f7752b5c7864b5f8a4ea07c47de9c666e9561390ca"
+    },
+    {
+      "bytes": 1425,
+      "path": "DATASET_CARD.md",
+      "sha256": "2984d35080d38afbbd02c4fc017fdb9146b74786330dfb16280d131c91c01c78"
+    },
+    {
+      "bytes": 663,
+      "path": "docs/limitations.md",
+      "sha256": "58903a354f70f4f3e602d84f450b4849ba9d00f51c1e3550314053125fcb8b64"
+    },
+    {
+      "bytes": 461,
+      "path": "docs/LIMITATIONS.md",
+      "sha256": "3428e3c5f62965c92e39e6517e19795b07a3adee2661e4acf5c9d56b8fed257c"
+    },
+    {
+      "bytes": 361,
+      "path": "docs/METHODOLOGY.md",
+      "sha256": "aab529fe46fbcb7eb83fd48e7e359e6d2bd10aa58661a6fa249e1f18ba547fe8"
+    },
+    {
+      "bytes": 1090,
+      "path": "docs/methods.md",
+      "sha256": "2835bde691eb62cdeedfc77ff2ed231fe74d424fedf0a7b96953b4a5b923f3f7"
+    },
+    {
+      "bytes": 283,
+      "path": "docs/REPRODUCTION.md",
+      "sha256": "1ba152f73c84ebb3ade2455ec38fff7b1c13709bb1c6e825569de3e6ff8d35c2"
+    },
+    {
+      "bytes": 306,
+      "path": "docs/SECURITY_AND_PRIVACY.md",
+      "sha256": "169a3738a8e4f4ceaf4388012b02111ffdefd6e7c915608caaa1949ff8b6912f"
+    },
+    {
+      "bytes": 170,
+      "path": "docs/suggested-tag.md",
+      "sha256": "519e311ce283705d6dcbbbb3ed89be15a1d94ebfd83f646f28a188b5da33208f"
+    },
+    {
+      "bytes": 997,
+      "path": "provenance/BUILD_PROVENANCE.json",
+      "sha256": "c1711b42e31a56809962cb2499009af23575434dc3f3fd4a272297f9aae47fbb"
+    },
+    {
+      "bytes": 10850,
+      "path": "provenance/LICENSE_AUDIT.json",
+      "sha256": "e9e3728b4e4e8b24b1a37cf8f2bf613cd0123f7e5f19b904f8ff5cf38ae9a616"
+    },
+    {
+      "bytes": 997,
+      "path": "provenance/PROVENANCE.json",
+      "sha256": "c1711b42e31a56809962cb2499009af23575434dc3f3fd4a272297f9aae47fbb"
+    },
+    {
+      "bytes": 394,
+      "path": "provenance/PUBLIC_EXPORT_LEAKAGE_CHECK.json",
+      "sha256": "e262a6495c3a7548822367e88d12268148d6458a84dc66a9b63d553c9f538d52"
+    },
+    {
+      "bytes": 4705,
+      "path": "provenance/PUBLIC_EXPORT_MANIFEST.json",
+      "sha256": "a2ec95e3493df2284c6fed1d4a7e2c510729794e64914095edba45731a30a4e4"
+    },
+    {
+      "bytes": 65,
+      "path": "provenance/PUBLIC_EXPORT_MANIFEST.sha256",
+      "sha256": "84a8f3f46de699c1c618c6d8fe46cf477b6ad0ad99636f782f1847afa16328b4"
+    },
+    {
+      "bytes": 1082,
+      "path": "provenance/PUBLIC_EXPORT_SANITIZATION_PROOF.json",
+      "sha256": "525960c306a747a64e425698676c392097135cbd7deb0744054070ad8b04af33"
+    },
+    {
+      "bytes": 704,
+      "path": "provenance/WITHHELD_COMPONENT_HASHES.json",
+      "sha256": "3f25c6a4b22cdcf8e3be858c6657cc238c3e1236aafd7d05eb68902518298230"
+    },
+    {
+      "bytes": 1847,
+      "path": "README.md",
+      "sha256": "f50d88be405179f70a6b8e6cefcf6b53acb072352c2d3cfe1404fb2c32344c0d"
+    },
+    {
+      "bytes": 1059,
+      "path": "RELEASE_NOTES.md",
+      "sha256": "53c234d4d6a898bd61d4d37758fca1f843b9a4715e511c5114727e580a77de97"
+    },
+    {
+      "bytes": 1217,
+      "path": "schema/generation-record.schema.json",
+      "sha256": "c5c3159dafd793ff9e646a11c4c1f23c8076eff17889a9acd4349c80700375c4"
+    },
+    {
+      "bytes": 441,
+      "path": "schema/public-export.schema.json",
+      "sha256": "dc8211ef99efe097b2de9efa593cdb867bed94b85963e027328d0ce26c9a2998"
+    },
+    {
+      "bytes": 615,
+      "path": "schema/task-record.schema.json",
+      "sha256": "68272ec4c35504fdeadd766bd0ce4fed44b9ea5cfedd2a895a4e14b3121d742c"
+    }
+  ],
+  "packageRoot": "publications/workledger-structured-edit-interface-study-v0.1",
+  "recordKind": "PUBLIC_EXPORT_MANIFEST",
+  "schemaVersion": 1,
+  "studyId": "workledger-structured-edit-interface-study-v0.1"
+}

provenance/PROVENANCE.json

@@ -0,0 +1,21 @@
+{
+  "generatedBy": "scripts/build-workledger-public-study-export.ts",
+  "offlineOnly": true,
+  "offlineReconciliationLabel": "OFFLINE_DERIVED_RECONCILIATION_NOT_A_REPLACEMENT_FOR_REGISTERED_ENDPOINT",
+  "providerRequests": 0,
+  "recordKind": "PUBLIC_EXPORT_BUILD_PROVENANCE",
+  "schemaVersion": 1,
+  "sourceArtifactLabels": [
+    "CALIBRATION_TASK_BANK_MANIFEST.json",
+    "AGGREGATE_ANALYSIS.json",
+    "PR48_CELL_STAGE_REPLAY.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/PER_MODEL_ANALYSIS.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/analysis/AGGREGATE_ANALYSIS.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/DECISION_CLASSIFICATION.json",
+    "qwen3-32b-v7-calibrated-multimodel-live-v1/cells/*/attempt-1/raw-response.txt"
+  ],
+  "sourceEvidenceTip": "c8e876133ed2745f23bc14d00aca79d68e83d590",
+  "sourceProtocolId": "qwen3-32b-v7-four-arm-interface-calibration-v1",
+  "studyId": "workledger-structured-edit-interface-study-v0.1",
+  "transportConstructed": false
+}

provenance/PUBLIC_EXPORT_LEAKAGE_CHECK.json

@@ -0,0 +1,15 @@
+{
+  "failClosed": true,
+  "findings": [],
+  "forbiddenKeyCount": 12,
+  "forbiddenStringPatternCount": 6,
+  "ok": true,
+  "recordKind": "PUBLIC_EXPORT_LEAKAGE_CHECK",
+  "scannedFileCount": 30,
+  "scannerExcludedFiles": [
+    "provenance/MANIFEST.json",
+    "provenance/PUBLIC_EXPORT_LEAKAGE_CHECK.json"
+  ],
+  "schemaVersion": 1,
+  "studyId": "workledger-structured-edit-interface-study-v0.1"
+}

provenance/PUBLIC_EXPORT_MANIFEST.json

@@ -0,0 +1,148 @@
+{
+  "entries": [
+    {
+      "bytes": 15058,
+      "path": "data/aggregate_results.jsonl",
+      "sha256": "81bf95b543e8d474c14c7b350978fce7bf541f0940c7720ee2ab89c40ff63c87"
+    },
+    {
+      "bytes": 20118,
+      "path": "data/aggregate-results.json",
+      "sha256": "6c8daea0e22c7aad2a93a37a533ebc571193ab01301e4eda329293469d57ca38"
+    },
+    {
+      "bytes": 10346,
+      "path": "data/claim_survival.jsonl",
+      "sha256": "3341108d1e51f0b1a249e7e7fdc9edce7fef59598e4817b4a1364964236a49c9"
+    },
+    {
+      "bytes": 11765,
+      "path": "data/claim-survival.json",
+      "sha256": "45aa24c6a3c0676fe888e6aa49893f4c05b42455e115342aa64cc3b738fa0c94"
+    },
+    {
+      "bytes": 667543,
+      "path": "data/generations.jsonl",
+      "sha256": "78363d7c143d289b092dde2c1ae1c674529d621573ff5c53d4c5618984b00299"
+    },
+    {
+      "bytes": 3177,
+      "path": "data/model-metadata.json",
+      "sha256": "16acadc395b0f211b65bae3ad89ea94675ec6645b16e73a6d6de11ad1bd8b628"
+    },
+    {
+      "bytes": 667543,
+      "path": "data/stage-outcomes.jsonl",
+      "sha256": "78363d7c143d289b092dde2c1ae1c674529d621573ff5c53d4c5618984b00299"
+    },
+    {
+      "bytes": 1165,
+      "path": "data/studies.jsonl",
+      "sha256": "80a2c92bc7657c91405f1fc6d89a33fe2b5e4547b22539fa83baadee90cbdc6d"
+    },
+    {
+      "bytes": 1165,
+      "path": "data/study_metadata.jsonl",
+      "sha256": "80a2c92bc7657c91405f1fc6d89a33fe2b5e4547b22539fa83baadee90cbdc6d"
+    },
+    {
+      "bytes": 46642,
+      "path": "data/tasks.jsonl",
+      "sha256": "416b1602fd0aedd2c34274f7752b5c7864b5f8a4ea07c47de9c666e9561390ca"
+    },
+    {
+      "bytes": 1425,
+      "path": "DATASET_CARD.md",
+      "sha256": "2984d35080d38afbbd02c4fc017fdb9146b74786330dfb16280d131c91c01c78"
+    },
+    {
+      "bytes": 663,
+      "path": "docs/limitations.md",
+      "sha256": "58903a354f70f4f3e602d84f450b4849ba9d00f51c1e3550314053125fcb8b64"
+    },
+    {
+      "bytes": 461,
+      "path": "docs/LIMITATIONS.md",
+      "sha256": "3428e3c5f62965c92e39e6517e19795b07a3adee2661e4acf5c9d56b8fed257c"
+    },
+    {
+      "bytes": 361,
+      "path": "docs/METHODOLOGY.md",
+      "sha256": "aab529fe46fbcb7eb83fd48e7e359e6d2bd10aa58661a6fa249e1f18ba547fe8"
+    },
+    {
+      "bytes": 1090,
+      "path": "docs/methods.md",
+      "sha256": "2835bde691eb62cdeedfc77ff2ed231fe74d424fedf0a7b96953b4a5b923f3f7"
+    },
+    {
+      "bytes": 283,
+      "path": "docs/REPRODUCTION.md",
+      "sha256": "1ba152f73c84ebb3ade2455ec38fff7b1c13709bb1c6e825569de3e6ff8d35c2"
+    },
+    {
+      "bytes": 306,
+      "path": "docs/SECURITY_AND_PRIVACY.md",
+      "sha256": "169a3738a8e4f4ceaf4388012b02111ffdefd6e7c915608caaa1949ff8b6912f"
+    },
+    {
+      "bytes": 170,
+      "path": "docs/suggested-tag.md",
+      "sha256": "519e311ce283705d6dcbbbb3ed89be15a1d94ebfd83f646f28a188b5da33208f"
+    },
+    {
+      "bytes": 997,
+      "path": "provenance/BUILD_PROVENANCE.json",
+      "sha256": "c1711b42e31a56809962cb2499009af23575434dc3f3fd4a272297f9aae47fbb"
+    },
+    {
+      "bytes": 10850,
+      "path": "provenance/LICENSE_AUDIT.json",
+      "sha256": "e9e3728b4e4e8b24b1a37cf8f2bf613cd0123f7e5f19b904f8ff5cf38ae9a616"
+    },
+    {
+      "bytes": 997,
+      "path": "provenance/PROVENANCE.json",
+      "sha256": "c1711b42e31a56809962cb2499009af23575434dc3f3fd4a272297f9aae47fbb"
+    },
+    {
+      "bytes": 1082,
+      "path": "provenance/PUBLIC_EXPORT_SANITIZATION_PROOF.json",
+      "sha256": "525960c306a747a64e425698676c392097135cbd7deb0744054070ad8b04af33"
+    },
+    {
+      "bytes": 704,
+      "path": "provenance/WITHHELD_COMPONENT_HASHES.json",
+      "sha256": "3f25c6a4b22cdcf8e3be858c6657cc238c3e1236aafd7d05eb68902518298230"
+    },
+    {
+      "bytes": 1847,
+      "path": "README.md",
+      "sha256": "f50d88be405179f70a6b8e6cefcf6b53acb072352c2d3cfe1404fb2c32344c0d"
+    },
+    {
+      "bytes": 1059,
+      "path": "RELEASE_NOTES.md",
+      "sha256": "53c234d4d6a898bd61d4d37758fca1f843b9a4715e511c5114727e580a77de97"
+    },
+    {
+      "bytes": 1217,
+      "path": "schema/generation-record.schema.json",
+      "sha256": "c5c3159dafd793ff9e646a11c4c1f23c8076eff17889a9acd4349c80700375c4"
+    },
+    {
+      "bytes": 441,
+      "path": "schema/public-export.schema.json",
+      "sha256": "dc8211ef99efe097b2de9efa593cdb867bed94b85963e027328d0ce26c9a2998"
+    },
+    {
+      "bytes": 615,
+      "path": "schema/task-record.schema.json",
+      "sha256": "68272ec4c35504fdeadd766bd0ce4fed44b9ea5cfedd2a895a4e14b3121d742c"
+    }
+  ],
+  "packageRoot": "publications/workledger-structured-edit-interface-study-v0.1",
+  "recordKind": "PUBLIC_EXPORT_MANIFEST_ALIAS",
+  "schemaVersion": 1,
+  "studyId": "workledger-structured-edit-interface-study-v0.1"
+}

provenance/PUBLIC_EXPORT_MANIFEST.sha256

@@ -0,0 +1 @@
+a2ec95e3493df2284c6fed1d4a7e2c510729794e64914095edba45731a30a4e4

provenance/PUBLIC_EXPORT_SANITIZATION_PROOF.json

@@ -0,0 +1,22 @@
+{
+  "credentialsExported": false,
+  "exportedFinalAssistantAnswersOnly": true,
+  "failClosed": true,
+  "generatedFromPrivateArtifactsByFieldSelection": true,
+  "hiddenEvaluatorSourceExported": false,
+  "multimodelAdapterPolicy": "Multimodel final answers are read from cells/*/attempt-1/raw-response.txt only when SHA-256 matches terminal evidence rawResponseHash; empty infrastructure failures emit empty answers only.",
+  "multimodelRawEvidenceExported": false,
+  "multimodelReasoningExported": false,
+  "multimodelRouteEvidenceExported": false,
+  "privatePathsExported": false,
+  "providerRequestsDuringExport": 0,
+  "rawProviderBodiesExported": false,
+  "reasoningExported": false,
+  "recordKind": "PUBLIC_EXPORT_SANITIZATION_PROOF",
+  "scannerScope": "All package files except provenance/MANIFEST.json and provenance/PUBLIC_EXPORT_LEAKAGE_CHECK.json.",
+  "schemaVersion": 1,
+  "sourceAnswerPolicy": "modelResponseText_or_raw-response.txt",
+  "sqliteExported": false,
+  "studyId": "workledger-structured-edit-interface-study-v0.1",
+  "transportConstructedDuringExport": false
+}

provenance/WITHHELD_COMPONENT_HASHES.json

@@ -0,0 +1,19 @@
+{
+  "hashesDisclosed": false,
+  "providerRequestsDuringExport": 0,
+  "reason": "Withheld components include hidden evaluator source and private transport evidence; this public package discloses only sanitized public records and non-sensitive manifest hashes.",
+  "recordKind": "PUBLIC_EXPORT_WITHHELD_COMPONENT_HASHES",
+  "schemaVersion": 1,
+  "studyId": "workledger-structured-edit-interface-study-v0.1",
+  "withheldComponents": [
+    "hidden evaluator source",
+    "private request bodies",
+    "raw provider response bodies",
+    "reasoning traces",
+    "authorization headers",
+    "idempotency keys",
+    "provider route evidence",
+    "response and request identifiers",
+    "SQLite ledgers"
+  ]
+}

schema/generation-record.schema.json

@@ -0,0 +1,61 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "additionalProperties": true,
+  "properties": {
+    "applicationOk": {
+      "type": "boolean"
+    },
+    "arm": {
+      "enum": [
+        "U0",
+        "U1",
+        "R",
+        "RN"
+      ]
+    },
+    "finalAssistantAnswer": {
+      "type": "string"
+    },
+    "finalAssistantAnswerSha256": {
+      "pattern": "^[a-f0-9]{64}$",
+      "type": "string"
+    },
+    "modelId": {
+      "type": "string"
+    },
+    "modelKey": {
+      "type": "string"
+    },
+    "recordKind": {
+      "enum": [
+        "PUBLIC_CALIBRATION_GENERATION_RECORD",
+        "PUBLIC_MULTIMODEL_GENERATION_RECORD"
+      ]
+    },
+    "registeredTerminalClassification": {
+      "type": "string"
+    },
+    "schemaVersion": {
+      "const": 1
+    },
+    "studyId": {
+      "const": "workledger-structured-edit-interface-study-v0.1"
+    },
+    "taskId": {
+      "type": "string"
+    }
+  },
+  "required": [
+    "schemaVersion",
+    "recordKind",
+    "studyId",
+    "taskId",
+    "arm",
+    "finalAssistantAnswer",
+    "finalAssistantAnswerSha256",
+    "registeredTerminalClassification",
+    "applicationOk"
+  ],
+  "title": "WorkLedger public generation record",
+  "type": "object"
+}

schema/public-export.schema.json

@@ -0,0 +1,21 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "additionalProperties": true,
+  "properties": {
+    "files": {
+      "items": {
+        "type": "string"
+      },
+      "type": "array"
+    },
+    "studyId": {
+      "const": "workledger-structured-edit-interface-study-v0.1"
+    }
+  },
+  "required": [
+    "studyId",
+    "files"
+  ],
+  "title": "WorkLedger structured edit interface study public export",
+  "type": "object"
+}

schema/task-record.schema.json

@@ -0,0 +1,30 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "additionalProperties": true,
+  "properties": {
+    "issueStatement": {
+      "type": "string"
+    },
+    "recordKind": {
+      "const": "PUBLIC_CALIBRATION_TASK_RECORD"
+    },
+    "schemaVersion": {
+      "const": 1
+    },
+    "studyId": {
+      "const": "workledger-structured-edit-interface-study-v0.1"
+    },
+    "taskId": {
+      "type": "string"
+    }
+  },
+  "required": [
+    "schemaVersion",
+    "recordKind",
+    "studyId",
+    "taskId",
+    "issueStatement"
+  ],
+  "title": "WorkLedger public calibration task record",
+  "type": "object"
+}