Mathematical Reasoning on MATH (Pass@1, TC)
90Pass@1 AccuracyEAPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EAPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 90 | 1.03 | |
| AEPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 89.1 | 1.01 | |
| ARPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 87.8 | 1.01 | |
| Reinforce++Backbone Model=Qwen2.5-7B-Instruct2026.06 | 87.6 | 1.12 | |
| ToolStarBackbone Model=Qwen2.5-7B-Instruct2026.06 | 86.4 | 1.28 | |
| GRPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 86 | 0.97 | |
| BaseBackbone Model=Qwen2.5-7B-Instruct2026.06 | 82 | — | |
| Reinforce++Backbone Model=Qwen2.5-3B-Instruct2026.06 | 81.8 | 1.12 | |
| EAPOBackbone Model=Qwen2.5-3B-Instruct2026.06 | 81.6 | 1.01 | |
| ARPOBackbone Model=Qwen2.5-3B-Instruct2026.06 | 80.8 | 1.05 | |
| GRPOBackbone Model=Qwen2.5-3B-Instruct2026.06 | 80.5 | 1.07 | |
| ToolStarBackbone Model=Qwen2.5-3B-Instruct2026.06 | 80.4 | 1.03 | |
| TIRBackbone Model=Qwen2.5-7B-Instruct2026.06 | 78.2 | 1.54 | |
| BaseBackbone Model=Qwen2.5-3B-Instruct2026.06 | 67.2 | — | |
| TIRBackbone Model=Qwen2.5-3B-Instruct2026.06 | 62.8 | 0.86 |