> Data revision: baf7c2f4307d3a989a8fb9363df8000fda3d9310
> Source files:
- [EVIDENCE-MAP.md](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/EVIDENCE-MAP.md)
- [results/publication-blockers.json](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/results/publication-blockers.json)

# Evidence map

Reviewed 2026-10-07. This map connects the spec clause inventory in the publication plan to the public benchmark record. It records what the current data can support and the measurement still needed. An implementation or conformance result shows parity with a stated contract; it does not show that the contract improves Build success.

## Reading this map

- **Empirical** means a claim about measured outcomes. The analysis status is shown beside each cited record.
- **Normative product** means a product rule or choice. It may be a current design decision without a comparative result.
- **Safety/correctness** means a behavior required for containment, integrity, or reliable state transitions. Conformance establishes implementation parity only.
- **Implementation freedom** means the spec leaves the implementation choice open or deferred.
- Claim IDs refer to [`results/claim-ledger.jsonl`](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/results/claim-ledger.jsonl). A row with status **INTERIM**, **CONFOUNDED**, **WITHDRAWN**, or **NOT-RUN** is not evidence of measured superiority.

The evidence map uses a locally sanitized v1.2 clause-reference snapshot in [spec/](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/spec/README.md). Its original cited content revision is 1118f7fc0dbb042af8c8de2ffd1b85768cf9e2a0; the upstream public sanitization revision is 30d9b25e4eefa7117121de44298fed504b429f4e. [spec/SOURCE.json](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/spec/SOURCE.json) records original source hashes separately from SHA-256 hashes of the published bytes. All 134 pinned v1.2 fragments linked below are locally verified against that snapshot. The decision ledger is absent, so comprehensive decision-to-clause coverage remains limited and cannot be independently verified.

Each clause page follows the authored clause map in `EVIDENCE-MAP.md`. Related IDs come from explicit source references and structured claim/crosswalk records.

- [1.1 Process model](/evidence/11-process-model/)
- [2.1 Intent .kogen/intents/&lt;slug&gt;/intent.md](/evidence/21-intent-kogenintentsintentmd/)
- [2.4 Acceptance tests and stack adapters](/evidence/24-acceptance-tests-and-stack-adapters/)
- [2.5 Approval, protection and refs](/evidence/25-approval-protection-and-refs/)
- [3.0 Rules](/evidence/30-rules/)
- [3.1 The recipe: ladder](/evidence/31-the-recipe-ladder/)
- [3.7 Verification and base-relative checks](/evidence/37-verification-and-base-relative-checks/)
- [3.8 Verdict, audit, selection](/evidence/38-verdict-audit-selection/)
- [3.9 Commit and landing](/evidence/39-commit-and-landing/)
- [4.1 Provider port and test seams](/evidence/41-provider-port-and-test-seams/)
- [4.9 Provider requests and prompt caching](/evidence/49-provider-requests-and-prompt-caching/)
- [4.10 Grok provider](/evidence/410-grok-provider/)
- [5.1 Process custody (success-relevant: deadlines, hangs, ARG_MAX)](/evidence/51-process-custody-success-relevant-deadlines-hangs-arg_max/)
- [6.1 Out of core v1](/evidence/61-out-of-core-v1/)
- [C.1 Definition](/evidence/c1-definition/)
- [Sources and precedence](/evidence/sources-and-precedence/)

## Decision-ID coverage

Every v1.2 clause fragment linked above is locally verified against the published snapshot. The source classification's section-level D references are present, but the underlying decision ledger is absent, so comprehensive decision-to-clause coverage remains limited and cannot be independently verified. Product decisions with no measured claim should retain an explicit **NORMATIVE** or **NO EMPIRICAL EVIDENCE** label.
