yinita/ps4mas-grpo-9b-fullft-smoke-best
017
PS4MAS GRPO full FT — Qwen3.5-9B (best)
- Run label:
grpo_9b_fullft_smoke_0624_0657 - Checkpoint kind: best
- Global step: 2
- Eval reward mean: -1.067
- Training state: see
training_state.json/checkpoint_best.jsonin run dir
Base model: Qwen/Qwen3.5-9B
