Refactor grader to use openenv.core.rubrics.WeightedSum + Rubric subclasses f0ca22d InosLihka commited on Apr 26
Add SKIP_EVAL flag to sft_on_hf.py for faster training-only runs ff20f02 InosLihka commited on Apr 26
Fix max_new_tokens for CoT format + add eval-only HF Jobs script b9c9b8f InosLihka commited on Apr 26
Algorithm Distillation: grader v2 with belief_accuracy + SFT pipeline ece0bbe InosLihka commited on Apr 26
docs: handoff bundle for new chat session + iter 4 partial analysis 7bb9278 InosLihka Claude Opus 4.7 (1M context) commited on Apr 26
tooling: scripts/analyze_iter.py + docs/results.md template d6d9e31 InosLihka Claude Opus 4.7 (1M context) commited on Apr 26
iter3: align reward with grader + belief-first format + exploration shaping 64d24b3 InosLihka Claude Opus 4.7 (1M context) commited on Apr 26
iter2: fix mode collapse + 3 deeper bugs from code review e21a960 InosLihka Claude Opus 4.7 (1M context) commited on Apr 25
tune: GRPO hyperparameter fixes from ML reviewer dc0186f InosLihka Claude Opus 4.7 (1M context) commited on Apr 25
feat: FAST_MODE preset for 10-15 min iteration cycles 1a865f8 InosLihka Claude Opus 4.7 (1M context) commited on Apr 25
env: meta-RL refactor (continuous profiles, action+belief, adaptation grader) ecbe0d8 InosLihka Claude Opus 4.7 (1M context) commited on Apr 25
InosLihka Claude Opus 4.7 (1M context) commited on 