Math Reasoning on BigMath level 5
43.1Average AccuracyDistillable
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DistillableBase Model=Gemma-3-12B2026.04 | 43.1 | 42.7 | 62.6 | |
| DistillableBase Model=Qwen3-8B2026.04 | 41.7 | 42 | 55.1 | |
| UndistillableBase Model=Gemma-3-12B2026.04 | 41.6 | 40.2 | 62.1 | |
| TeacherBase Model=Qwen3-8B2026.04 | 39.4 | 39.2 | 52 | |
| TeacherBase Model=Gemma-3-12B2026.04 | 39.3 | 38.1 | 61.4 | |
| UndistillableBase Model=Qwen3-8B2026.04 | 38.7 | 38.1 | 52.6 | |
| ImproveBase Model=Qwen3-1.7B2026.04 | 37.9 | 37.8 | 54.7 | |
| ImproveBase Model=Gemma-3-4B2026.04 | 28 | 27.4 | 51 | |
| StudentBase Model=Qwen3-1.7B2026.04 | 27 | 21.9 | 45.4 | |
| GKD-RKLBase Model=Qwen3-1.7B2026.04 | 25.9 | 25.4 | 46.2 | |
| GKD-RKLBase Model=Gemma-3-4B2026.04 | 25.5 | 24 | 50.5 | |
| SFTBase Model=Gemma-3-4B2026.04 | 24 | 22.5 | 49 | |
| GKD-FKLBase Model=Qwen3-1.7B2026.04 | 23.7 | 22.3 | 46.1 | |
| GKD-FKLBase Model=Gemma-3-4B2026.04 | 23.6 | 21.2 | 49.9 | |
| SFTBase Model=Qwen3-1.7B2026.04 | 22.2 | 20.7 | 44.5 | |
| StudentBase Model=Gemma-3-4B2026.04 | 21.6 | 19.6 | 46.4 | |
| MisledBase Model=Qwen3-1.7B2026.04 | 9.1 | 5.7 | 23.7 | |
| MisledBase Model=Gemma-3-4B2026.04 | 6.3 | 2 | 25.5 |