logan7000/cogrpo-homo-qwen25-3b-math345-groupA
029
archive: wandb curve + provenance (run ujg2tvly)
Add cogrpo-homo-qwen25-3b-math345-groupA best ckpt (eval_reward 0.666@s110)
initial commit
archive: wandb curve + provenance (run ujg2tvly)
Add cogrpo-homo-qwen25-3b-math345-groupA best ckpt (eval_reward 0.666@s110)
initial commit