Commit History
8 days ago
-
Regrade all 4,650 CrossSearch answers with GPT-6.1 Sol
Add versioned per-rubric grades and recomputed current metrics. Preserve original generation traces, original judgments, historical scores, private visibility and evaluation-only canaries.a79b199
19 days ago
-
Add 100-task resources and first rollout shards
b54a239