Mathematical Reasoning on MATH500 (Pass@1, TC)
78.8Pass@1EAPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EAPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 78.8 | 0.98 | |
| ARPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 78 | 1.02 | |
| GRPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 77.6 | 0.98 | |
| AEPOBackbone Model=Qwen2.5-7B-Instruct2026.06 | 77.6 | 1.1 | |
| Reinforce++Backbone Model=Qwen2.5-7B-Instruct2026.06 | 76.8 | 1.02 | |
| ToolStarBackbone Model=Qwen2.5-7B-Instruct2026.06 | 76.8 | 1.38 | |
| BaseBackbone Model=Qwen2.5-7B-Instruct2026.06 | 70.6 | — | |
| EAPOBackbone Model=Qwen2.5-3B-Instruct2026.06 | 70.1 | 1.03 | |
| GRPOBackbone Model=Qwen2.5-3B-Instruct2026.06 | 70 | 1.14 | |
| ARPOBackbone Model=Qwen2.5-3B-Instruct2026.06 | 69.8 | 1.07 | |
| Reinforce++Backbone Model=Qwen2.5-3B-Instruct2026.06 | 69.4 | 1.09 | |
| ToolStarBackbone Model=Qwen2.5-3B-Instruct2026.06 | 69 | 1.01 | |
| TIRBackbone Model=Qwen2.5-7B-Instruct2026.06 | 68.2 | 0.51 | |
| BaseBackbone Model=Qwen2.5-3B-Instruct2026.06 | 59.2 | — | |
| TIRBackbone Model=Qwen2.5-3B-Instruct2026.06 | 52.2 | 0.81 |