Mathematical Reasoning on MATH 500 (Pass@32)
0.941Pass@32OPEFO
Evaluation Results
| Method | Links | |
|---|---|---|
| OPEFOBackbone=Qwen2.5-Math-7B2026.05 | 0.941 | |
| KL-CovBackbone=Qwen2.5-Math-7B2026.05 | 0.937 | |
| Clip-CovBackbone=Qwen2.5-Math-7B2026.05 | 0.935 | |
| GRPO (Strict)Backbone=Qwen2.5-Math-7B2026.05 | 0.933 | |
| Clip-higherBackbone=Qwen2.5-Math-7B2026.05 | 0.927 | |
| GRPOBackbone=Qwen2.5-Math-7B2026.05 | 0.925 | |
| Entropy-RegBackbone=Qwen2.5-Math-7B2026.05 | 0.92 |