Mathematical Reasoning on AMC 23 (Avg@16, P@16)
46.72Avg@16PriFT-mass
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PriFT-massModel=Qwen2.5-Math (1.5B)2026.06 | 46.72 | 92.5 | |
| PriFT-probModel=Qwen2.5-Math (1.5B)2026.06 | 38.91 | 87.5 | |
| TALRModel=Qwen2.5-Math (1.5B)2026.06 | 38.75 | 77.5 | |
| DFTModel=Qwen2.5-Math (1.5B)2026.06 | 38.13 | 70 | |
| ASFTModel=Qwen2.5-Math (1.5B)2026.06 | 32.97 | 80 | |
| IDFTModel=Qwen2.5-Math (1.5B)2026.06 | 31.72 | 72.5 | |
| PriFT-massModel=Qwen2.5-Instruct (1.5B)2026.06 | 27.81 | 67.5 | |
| PriFT-probModel=Qwen2.5-Instruct (1.5B)2026.06 | 25.47 | 75 | |
| PriFT-massModel=DeepSeekMath (7B)2026.06 | 21.72 | 60 | |
| DFTModel=Qwen2.5-Instruct (1.5B)2026.06 | 20.94 | 60 | |
| DFTModel=DeepSeekMath (7B)2026.06 | 20.47 | 57.5 | |
| PriFT-probModel=DeepSeekMath (7B)2026.06 | 19.69 | 65 | |
| SFTModel=Qwen2.5-Math (1.5B)2026.06 | 19.38 | 70 | |
| EAFTModel=Qwen2.5-Math (1.5B)2026.06 | 18.59 | 70 | |
| PriFT-massModel=Qwen2.5 (1.5B)2026.06 | 18.59 | 65 | |
| PretrainedModel=Qwen2.5-Instruct (1.5B)2026.06 | 18.59 | 67.5 | |
| DFTModel=Qwen2.5 (1.5B)2026.06 | 17.97 | 62.5 | |
| OriginalModel=Qwen2.5-Math (1.5B)2026.06 | 15.94 | 82.5 | |
| PriFT-probModel=Qwen2.5 (1.5B)2026.06 | 14.84 | 62.5 | |
| PretrainedModel=DeepSeekMath (7B)2026.06 | 14.38 | 62.5 | |
| SFTModel=DeepSeekMath (7B)2026.06 | 12.03 | 55 | |
| SFTModel=Qwen2.5-Instruct (1.5B)2026.06 | 8.44 | 62.5 | |
| SFTModel=Qwen2.5 (1.5B)2026.06 | 7.81 | 47.5 | |
| PretrainedModel=Qwen2.5 (1.5B)2026.06 | 2.5 | 22.5 |