Mathematical Reasoning on Gaokao En 23
78.4AccuracyGPT-o1-mini
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-o1-miniTool=✗2026.02 | 78.4 | |
| LaPha / sg@128Base model=Qwen2.5-Math-7B, Tool=✓2026.02 | 74.3 | |
| LaPha / sg@128Base model=Qwen2.5-7B, Tool=✓2026.02 | 71.4 | |
| LaPha / sg@128Base model=Qwen2.5-Math-1.5B, Tool=✓2026.02 | 67.7 | |
| GPT-4oTool=✗2026.02 | 67.5 | |
| DAPOBase model=Qwen2.5-7B, Tool=✗2026.02 | 63.2 | |
| DAPOBase model=Qwen2.5-Math-1.5B, Tool=✗2026.02 | 59.1 | |
| LaPhaBase model=Qwen2.5-Math-1.5B, Tool=✓2026.02 | 58.8 | |
| SFTBase model=Qwen2.5-Math-1.5B, Tool=✓2026.02 | 53.7 | |
| LaPha / sg@128Base model=Qwen2.5-1.5B, Tool=✓2026.02 | 50.1 | |
| SFTBase model=Qwen2.5-Math-1.5B, Tool=✗2026.02 | 46.5 | |
| HybKDTeacher-Student Pair=Qwen2.5-Math-7B → 1.5B2026.05 | 44.1 | |
| DAPOBase model=Qwen2.5-Math-7B, Tool=✗2026.02 | 44 | |
| LaPhaBase model=Qwen2.5-1.5B, Tool=✓2026.02 | 43.4 | |
| α-β divergenceTeacher-Student Pair=Qwen2.5-Math-7B → 1.5B2026.05 | 42.88 | |
| DAPOBase model=Qwen2.5-1.5B, Tool=✗2026.02 | 42.3 | |
| Skew F(R)KLTeacher-Student Pair=Qwen2.5-Math-7B → 1.5B2026.05 | 42.23 | |
| Forward KLTeacher-Student Pair=Qwen2.5-Math-7B → 1.5B2026.05 | 41.77 | |
| Hard KDTeacher-Student Pair=Qwen2.5-Math-7B → 1.5B2026.05 | 41.04 | |
| SFTBase model=Qwen2.5-Math-7B, Tool=✓2026.02 | 39 | |
| Reverse KLTeacher-Student Pair=Qwen2.5-Math-7B → 1.5B2026.05 | 37.35 | |
| SFTBase model=Qwen2.5-Math-7B, Tool=✗2026.02 | 37.1 | |
| SFTBase model=Qwen2.5-7B, Tool=✗2026.02 | 36.4 | |
| SFTBase model=Qwen2.5-1.5B, Tool=✗2026.02 | 19.7 |