benchmarks: construct an order-12 Hadamard certificate - #593
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6979ddd5de
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
6979ddd to
29a5385
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 29a5385c66
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a61199ea22
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
- Change evidence media type from text/plain to application/json with payload_shape in instruction.md and public_contract.json - Decouple mathematics, evidence, scope, and assurance from contract; use require_input_binding=False and independent computation - Emit canonical reward dimensions (correctness, evidence_validity, scope_accuracy, assurance_calibration) in both main path and exception handler - Bound evidence memory to 2 MiB via max_bytes in evidence_list_is_bound and read_evidence_json - Accept all scoreable assurances (UNVERIFIED, COMPUTED, CHECKED) - Drop whitespace-churn commits during rebase onto origin/main
… lint Co-authored-by: morluto <morluto@users.noreply.github.com>
488345c to
18edbc3
Compare
3 participants
Summary
C-007Benchmark contract
Shortcut and duplication audit
Knowing that order 12 exists or naming Paley is insufficient. The full matrix, full Gram matrix, normalization, and determinant are independently checked. The verifier accepts row/column-equivalent normalized matrices. This does not duplicate the bounded Vizing graph workflow or the perfect-cuboid semantic-scope audit.
Validation
git diff --check--network noneNative Harbor 0.20 on macOS remains blocked by its Docker no-network capability detection before task execution. Network policy was not weakened and no Docker socket was mounted. This draft awaits the GitHub Linux exact-task Oracle before completion.