2 Commits

Author	SHA1	Message	Date
transcrilive	81e8ac88cc	feat(config): add enable_thinking flag (default False) + fix HMMT bench gold answers ... Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-10 13:08:41 +02:00
transcrilive	6745416228	feat(bench): HMMT/AIME small-subset harness + answer extraction tests	2026-05-10 03:20:33 +02:00