Problem Solving on Qwen 7B 2.5
22.2End Pass@1REINFORCE + ES
Evaluation Results
| Method | Links | |
|---|---|---|
| REINFORCE + ESRL update=REINFORCE, Orchestration=ES, n=32026.06 | 22.2 | |
| GRPO + none (A0)RL update=GRPO, Orchestration=none (A0), n=52026.06 | 20.7 | |
| GRPO + CARE (A3)RL update=GRPO, Orchestration=CARE (A3), n=52026.06 | 20.4 | |
| GRPO + ESRL update=GRPO, Orchestration=ES, n=32026.06 | 17 | |
| REINFORCE + CARE (A3)RL update=REINFORCE, Orchestration=CARE (A3), n=52026.06 | 13.8 | |
| REINFORCE + none (A0)RL update=REINFORCE, Orchestration=none (A0), n=52026.06 | 11.8 |