Mathematical Reasoning on MATH 500 (Score)
79.95ScoreMathForge
Evaluation Results
| Method | Links | |
|---|---|---|
| MathForgeBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 79.95 | |
| MQRBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 77.85 | |
| DGPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 74.25 | |
| GRPO-ADBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 73.25 | |
| GSPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 73 | |
| DAPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 72.7 | |
| GRPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 72.2 | |
| Dr.GRPOBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 72.05 | |
| GPGBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 72.05 | |
| Base ModelBackbone=Qwen2.5-Math-7B, Training Dataset=MATH, Evaluation Protocol=Zero-shot2026.01 | 48.6 |