CoolFace
Modelpublic

lucabaroni/gpt-oss-20b-rlvr-no-conftest-20260909-completion

sourceHugging Faceapache-2.0updated 18d agoView on Hugging Face
0likes18downloads
Model Card

gptoss20b — final

Rank-32 adapter from matched no-conftest RLVR.

See study_provenance.json for the exact base revision, training configuration and checkpoint lineage. Base model license: https://huggingface.co/openai/gpt-oss-20b/blob/6cee5e81ee83917806bbde320786a8fb61efebee/README.md

Load with PEFT using the pinned base revision and this adapter revision. The adapter contains learned evaluator exploitation in a prompted experimental environment.