Commit graph

7 commits

Author SHA1 Message Date
Shay
d34883b61a feat(evals): implement safe metadata and fail-closed evaluator check for GSM1K 2026-06-23 07:52:32 -07:00
Shay
db081fa5b3 feat(evals): add GSM1K local audit adapter 2026-06-23 07:40:53 -07:00
Shay
f823eb437b feat(evals): add generalization audit runner skeleton 2026-06-23 07:23:43 -07:00
Shay
ed8c0b0e43 feat(evals): verify local generalization benchmark cache 2026-06-23 07:03:56 -07:00
Shay
83ba9a2e9b feat(evals): implement manifest schema and policy linter (#886) 2026-06-23 06:30:12 -07:00
Shay
67f9b63916 fix(evals): align benchmark manifests and gitignore with sealed-holdout policy
- Adds benchmark cache ignore rule to extend ADR-0119.7 discipline
- Removes sealed slice references from manifest smoke fixture lines
- Clarifies openbookqa and para_mawps license gates
2026-06-23 06:12:50 -07:00
Shay
34dc0d64c5 feat(evals): add generalization benchmark manifests + policy (PR-1 of 2)
Adds policy document and sealed manifest records for the first 8
external audit datasets. No data is vendored. Local cache paths are
gitignored. Fetch/verify scripts and smoke fixtures come in PR-2.

Datasets: GSM1K, ASDiv, SVAMP, PARA-MAWPS, ARC-Easy, ARC-Challenge,
          OpenBookQA, CLUTRR
2026-06-23 05:59:36 -07:00