Mathematical Reasoning on MATH500 (Pass@1 and Pass@3)
85.6Pass@1UPFT (2025)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UPFT (2025)Backbone=Qwen3-8B2026.01 | 85.6 | 92.4 | |
| P-ALIGNBackbone=Qwen3-8B2026.01 | 84.8 | 92.8 | |
| SFT (Long-CoT)Backbone=Qwen3-8B2026.01 | 84.4 | 90.8 | |
| Zero-ShotBackbone=Qwen3-8B2026.01 | 83.6 | 91.2 | |
| P-ALIGNBackbone=Qwen2.5-7B-Instruct2026.01 | 75.8 | 85.2 | |
| SFT (Long-CoT)Backbone=Qwen2.5-7B-Instruct2026.01 | 75.6 | 86.6 | |
| UPFT (2025)Backbone=Qwen2.5-7B-Instruct2026.01 | 74.8 | 84.6 | |
| SFT (Label)Backbone=Qwen3-8B2026.01 | 73.2 | 84.6 | |
| Zero-ShotBackbone=Qwen2.5-7B-Instruct2026.01 | 69.8 | 84.8 | |
| SFT (Label)Backbone=Qwen2.5-7B-Instruct2026.01 | 62.2 | 71 |