Mathematical Reasoning on GSM8K (Pass@1, Avg.)
92.27Pass@1+CurES-GRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| +CurES-GRPOBase Model=Qwen2.5-Math-7B, Training Method=CurES-GRPO2025.10 | 92.27 | 52.41 | |
| +RPPBase Model=Qwen2.5-Math-7B, Training Method=RPP2025.10 | 91.89 | 50.18 | |
| +CurES-RPPBase Model=Qwen2.5-Math-7B, Training Method=CurES-RPP2025.10 | 91.89 | 49.62 | |
| +Speed-RL-RPPBase Model=Qwen2.5-Math-7B, Training Method=Speed-RL-RPP2025.10 | 91.81 | 49.56 | |
| +GRPOBase Model=Qwen2.5-Math-7B, Training Method=GRPO2025.10 | 91.43 | 47.59 | |
| +GVM-GRPOBase Model=Qwen2.5-Math-7B, Training Method=GVM-GRPO2025.10 | 91.28 | 50.64 | |
| +GVM-RPPBase Model=Qwen2.5-Math-7B, Training Method=GVM-RPP2025.10 | 90.07 | 46.95 | |
| +Speed-RL-GRPOBase Model=Qwen2.5-Math-7B, Training Method=Speed-RL-GRPO2025.10 | 88.7 | 48.84 | |
| +GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=GRPO2025.10 | 86.43 | 41.64 | |
| +CurES-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=CurES-GRPO2025.10 | 85.97 | 44.94 | |
| +CurES-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=CurES-RPP2025.10 | 85.82 | 44.14 | |
| +Speed-RL-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=Speed-RL-GRPO2025.10 | 85.67 | 41.34 | |
| +Speed-RL-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=Speed-RL-RPP2025.10 | 85.67 | 41.75 | |
| +GVM-GRPOBase Model=Qwen2.5-Math-1.5B, Training Method=GVM-GRPO2025.10 | 84.23 | 42.82 | |
| +GVM-RPPBase Model=Qwen2.5-Math-1.5B, Training Method=GVM-RPP2025.10 | 84 | 42.21 | |
| +RPPBase Model=Qwen2.5-Math-1.5B, Training Method=RPP2025.10 | 82.94 | 35.86 | |
| Qwen2.5-Math-7BBase Model=Qwen2.5-Math-7B, Training Method=Base2025.10 | 72.4 | 33.48 | |
| Qwen2.5-Math-1.5BBase Model=Qwen2.5-Math-1.5B, Training Method=Base2025.10 | 43.9 | 20 |