Mathematical Reasoning on MATH (test) (Pass@k Suite)
77.2Pass@1AGPO
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| AGPOBackbone=Qwen2.5-Math-7B2026.05 | 77.2 | 83.5 | 87.9 | 90.7 | 92.7 | 94.3 | 95.6 | 96.4 | 96.6 | |
| W-REINFORCEBackbone=Qwen2.5-Math-7B2026.05 | 76.3 | 82.4 | 86.9 | 90.2 | 92.3 | 93.8 | 95.4 | 96.2 | 96.6 | |
| GRPOBackbone=Qwen2.5-Math-7B2026.05 | 76.2 | 81.5 | 85.2 | 88 | 90 | 91.7 | 93 | 94 | 94.7 | |
| PPOBackbone=Qwen2.5-Math-7B2026.05 | 75.8 | 81.5 | 85.5 | 88.4 | 90.6 | 92.2 | 93.4 | 94.3 | 94.8 | |
| REINFORCEBackbone=Qwen2.5-Math-7B2026.05 | 74.1 | 78.7 | 81.9 | 84.4 | 86.8 | 88.2 | 89.6 | 91 | 92.3 | |
| Base ModelBackbone=Qwen2.5-Math-7B2026.05 | 63.1 | 76.2 | 83.5 | 88.1 | 91.5 | 93.6 | 95.3 | 96.2 | 96.7 |