Mathematical Reasoning on AMC 23 (Avg@8, Pass@8, Average Response Length)
31.12Avg@8PowerOPD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PowerOPDalpha=5, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 31.12 | 57.5 | 2,456 | |
| PowerOPDalpha=10, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 29.69 | 57.5 | 2,426 | |
| PowerOPDalpha=500, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 29.25 | 57.5 | 2,049 | |
| Full-vocabTeacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 28.75 | 40 | 4,096 | |
| PowerOPDalpha=0.1, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 28.12 | 57.5 | 2,558 | |
| PowerOPDalpha=100, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 28.12 | 57 | 2,230 | |
| PowerOPDalpha=0.5, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 26.88 | 55 | 2,617 | |
| PowerOPDalpha=50, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 24.69 | 60 | 2,358 | |
| PowerOPDalpha=1, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 24.38 | 47.5 | 2,628 | |
| VanillaParam=+ Tanh, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 22.69 | 52.5 | 2,609 | |
| VanillaParam=+ Clip, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 21.88 | 52.5 | 2,669 | |
| VanillaTeacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 14.37 | 47.5 | 4,096 | |
| VanillaParam=+ Z-Score, Teacher=Qwen3-4B, Student=Qwen3-0.6B-Base2026.06 | 14.37 | 45 | 2,648 |