Mathematical Reasoning on AIME 24/25 (Pass@32, Avg)
14.3Average Score (Pass@32)OPDfull-V
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OPDfull-VModel Setup=Qwen3-4B → Qwen3-4B-Base2026.05 | 14.3 | 40 | 45.1 | |
| vOPDfull-VModel Setup=Qwen3-4B → Qwen3-4B-Base2026.05 | 13.5 | 40 | 45.4 | |
| vOPDtop-kModel Setup=Qwen3-4B → Qwen3-4B-Base2026.05 | 13.2 | 41.7 | 45.3 | |
| OPDModel Setup=Qwen3-4B → Qwen3-4B-Base2026.05 | 10.4 | 38.3 | 42.9 | |
| OPDtop-kModel Setup=Qwen3-4B → Qwen3-4B-Base2026.05 | 10.1 | 41.7 | 42.8 | |
| OPDfull-VModel Setup=Olmo-3-7B-Think → Olmo-3-7B-Base2026.05 | 8.6 | 23.4 | 31.8 | |
| vOPDtop-kModel Setup=Olmo-3-7B-Think → Olmo-3-7B-Base2026.05 | 8.4 | 25 | 33.1 | |
| vOPDfull-VModel Setup=Olmo-3-7B-Think → Olmo-3-7B-Base2026.05 | 7.9 | 21.7 | 32.9 | |
| StudentModel Setup=Qwen3-4B → Qwen3-4B-Base2026.05 | 7.8 | 26.8 | 28.4 | |
| OPDModel Setup=Olmo-3-7B-Think → Olmo-3-7B-Base2026.05 | 7.4 | 21.7 | 29.9 | |
| vOPDfull-VModel Setup=Qwen3-1.7B → Qwen3-1.7B-Base2026.05 | 6.2 | 25 | 33.1 | |
| OPDfull-VModel Setup=Qwen3-1.7B → Qwen3-1.7B-Base2026.05 | 5.8 | 25 | 33 | |
| vOPDtop-kModel Setup=Qwen3-1.7B → Qwen3-1.7B-Base2026.05 | 5.6 | 26.7 | 33 | |
| OPDtop-kModel Setup=Olmo-3-7B-Think → Olmo-3-7B-Base2026.05 | 5.5 | 15 | 29.5 | |
| StudentModel Setup=Olmo-3-7B-Think → Olmo-3-7B-Base2026.05 | 5.4 | 30 | 20.2 | |
| OPDModel Setup=Qwen3-1.7B → Qwen3-1.7B-Base2026.05 | 4.8 | 28.3 | 29.8 | |
| OPDtop-kModel Setup=Qwen3-1.7B → Qwen3-1.7B-Base2026.05 | 4 | 30 | 30.2 | |
| StudentModel Setup=Qwen3-1.7B → Qwen3-1.7B-Base2026.05 | 3 | 20 | 20.6 |