core/evals/gsm8k_math/holdout_dev
Shay 471d6721e2 feat(reader): integrate mass-noun comparative logic for case 0148
Commits the WIP parser integration and recognizer logic left by previous session. Updates graph logic, refusal taxonomy filters, roundtrip token matching, and recognizer_match patterns to formally support mass-noun comparatives.
Passes exact holdout_dev/v1 261-case measurement and full smoke validation.
2026-07-19 11:40:37 -07:00
..
v1 feat(reader): integrate mass-noun comparative logic for case 0148 2026-07-19 11:40:37 -07:00
__init__.py feat(eval): held-out dev lane — the honest iteration metric (real capability = 0) 2026-06-04 02:30:42 -07:00