CoolFace
Modelpublic

SelectiveDOPD/JustRL-Qwen3-4b-RKLSelective-Top10pct

sourceHugging Faceupdated 15d agoView on Hugging Face
0likes320downloads
README.md26 linesDownload Raw Back to root
1---2library_name: transformers3pipeline_tag: text-generation4---5 6# JustRL-Qwen3-4b-RKLSelective-Top10pct7 8Uploaded from `justrl_qwen3_4b_rkl_ladder_90_100_kl` in the BiDirect-OPD experiments.9 10The `main` branch contains `global_step_300`. Available earlier checkpoint branches:11 12- `global_step_20`13- `global_step_40`14- `global_step_60`15- `global_step_80`16- `global_step_100`17- `global_step_120`18- `global_step_140`19- `global_step_160`20- `global_step_180`21- `global_step_200`22- `global_step_220`23- `global_step_240`24- `global_step_260`25- `global_step_280`26