Mathematical Reasoning on AMC 23 (Avg@32)
79.3Avg@32MHPO
Evaluation results
| Method | Links | |
|---|---|---|
| MHPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 79.3 | |
| MHPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 79.3 | |
| DAPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 75.8 | |
| DAPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 75.8 | |
| MHPOBackbone=Qwen2.5-7B-Instruct2026.03 | 74.6 | |
| GSPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 74.6 | |
| GSPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 74.6 | |
| SAPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 72.5 | |
| SAPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 72.5 | |
| GRPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 72.4 | |
| GRPOBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 72.4 | |
| STEERBackbone=Qwen2.5-Math-7B2025.10 | 72.1 | |
| OPOBackbone=Qwen2.5-Math-7B2025.10 | 71.5 | |
| DAPOBackbone=Qwen2.5-7B-Instruct2026.03 | 71.3 | |
| Entro. Adv.Backbone=Qwen2.5-Math-7B2025.10 | 70.2 | |
| GSPOBackbone=Qwen2.5-7B-Instruct2026.03 | 70.1 | |
| SAPOBackbone=Qwen2.5-7B-Instruct2026.03 | 69.2 | |
| Clip–CovBackbone=Qwen2.5-Math-7B2025.10 | 68.4 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.03 | 68.3 | |
| GRPO w/ Entro. LossBackbone=Qwen2.5-Math-7B2025.10 | 67.6 | |
| GRPO w/ clip-highBackbone=Qwen2.5-Math-7B2025.10 | 66.8 | |
| GRPOBackbone=Qwen2.5-Math-7B2025.10 | 66.2 | |
| GRPO w/ Fork TokensBackbone=Qwen2.5-Math-7B2025.10 | 65.5 | |
| SimpleRL-ZooBackbone=Qwen2.5-Math-7B2025.10 | 65.4 | |
| Eurus-PRIMEBackbone=Qwen2.5-Math-7B2025.10 | 65.2 | |
| W-REINFORCEBackbone=Qwen2.5-Math-7B2025.10 | 64.9 | |
| KL–CovBackbone=Qwen2.5-Math-7B2025.10 | 64.2 | |
| BaselineBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 56.3 | |
| BaselineBackbone=Qwen2.5-Math-7B-Instruct2026.03 | 56.3 | |
| BaselineBackbone=Qwen2.5-7B-Instruct2026.03 | 52.4 | |
| Qwen2.5-Math-7BBackbone=Qwen2.5-Math-7B2025.10 | 44.6 |