Math Reasoning on BigMath level 4
66.3AccuracyDistillable
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DistillableBase Model=Qwen3-8B2026.04 | 66.3 | 66.4 | 76.3 | |
| DistillableBase Model=Gemma-3-12B2026.04 | 65.6 | 66.4 | 80.3 | |
| UndistillableBase Model=Qwen3-8B2026.04 | 65.1 | 66.2 | 74.8 | |
| UndistillableBase Model=Gemma-3-12B2026.04 | 64.9 | 64.9 | 79.4 | |
| TeacherBase Model=Gemma-3-12B2026.04 | 63.9 | 65.2 | 79.3 | |
| TeacherBase Model=Qwen3-8B2026.04 | 62.5 | 62.6 | 72.5 | |
| ImproveBase Model=Qwen3-1.7B2026.04 | 61.5 | 61.2 | 77.3 | |
| StudentBase Model=Qwen3-1.7B2026.04 | 52.6 | 56.5 | 69 | |
| ImproveBase Model=Gemma-3-4B2026.04 | 52.3 | 51.8 | 76.3 | |
| GKD-RKLBase Model=Qwen3-1.7B2026.04 | 50.2 | 50.9 | 70.2 | |
| GKD-RKLBase Model=Gemma-3-4B2026.04 | 49.5 | 48.8 | 76.5 | |
| SFTBase Model=Gemma-3-4B2026.04 | 48.5 | 47.5 | 76 | |
| GKD-FKLBase Model=Gemma-3-4B2026.04 | 48.2 | 47.4 | 75.7 | |
| GKD-FKLBase Model=Qwen3-1.7B2026.04 | 46.5 | 47.9 | 70.8 | |
| SFTBase Model=Qwen3-1.7B2026.04 | 46 | 46.8 | 70.9 | |
| StudentBase Model=Gemma-3-4B2026.04 | 41.3 | 39.9 | 67.4 | |
| MisledBase Model=Qwen3-1.7B2026.04 | 21.5 | 16.7 | 38 | |
| MisledBase Model=Gemma-3-4B2026.04 | 16.5 | 8 | 33.9 |