Mathematical Reasoning on AMC23 (Score)
61.02ScoreDGPO
Evaluation results
| Method | Links | |
|---|---|---|
| DGPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 61.02 | |
| GSPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 60.16 | |
| MathForgeBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 59.84 | |
| GPGBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 59.61 | |
| MQRBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 59.38 | |
| GRPO-ADBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 59.06 | |
| GRPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 58.98 | |
| Dr.GRPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 58.59 | |
| DAPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 58.2 | |
| Base ModelBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 35.23 |