Mathematical Reasoning on MATH 500 (Avg@32)
85.2Average Success Rate @32MHPO
Evaluation results
| Method | Links | |
|---|---|---|
| MHPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 85.2 | |
| DAPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 84.4 | |
| GSPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 83.8 | |
| SAPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 83.6 | |
| GRPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 82.3 | |
| GRPOModel=Qwen2.5-Math-7B, G=22026.06 | 77.43 | |
| GRPOModel=Qwen2.5-Math-7B, G=162026.06 | 76.94 | |
| C-RF w/ NTFModel=Qwen2.5-Math-7B, G=12026.06 | 74.08 | |
| BaselineBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 73.1 | |
| RF++Model=Qwen2.5-Math-7B, G=82026.06 | 71.53 | |
| GRPOModel=Qwen2.5-Math-1.5B, G=162026.06 | 71.04 | |
| C-RF w/ NTFModel=Qwen2.5-Math-1.5B, G=12026.06 | 69.63 | |
| GRPOModel=Qwen2.5-Math-1.5B, G=22026.06 | 68.81 | |
| RF++ w/ NTFModel=Qwen2.5-Math-1.5B, G=12026.06 | 68.21 | |
| RF++Model=Qwen2.5-Math-1.5B, G=82026.06 | 67.48 | |
| RF++ w/ NTFModel=Qwen2.5-Math-7B, G=12026.06 | 64.94 |