Mathematical Reasoning on Mathematical Reasoning Average
53.4Pass@1AOPD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| AOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 53.4 | 67.43 | |
| GKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 49.65 | 66.96 | |
| OPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 49.31 | 66.64 | |
| AOPDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 49.16 | 64.07 | |
| ExOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 48.2 | 64.53 | |
| ExOPDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 47.5 | 62.63 | |
| OPDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 47.29 | 64.15 | |
| GKDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 46.11 | 63.71 | |
| SFT CheckpointBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 45.9 | 61.96 | |
| AOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 43.13 | 63.66 | |
| SFT CheckpointBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 41.87 | 58.56 | |
| GKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 40.9 | 60.61 | |
| SeqKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 38.82 | 59.11 | |
| SFT CheckpointBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 36.53 | 53.86 | |
| SeqKDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 35.69 | 54.44 | |
| ExOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 35.41 | 54.73 | |
| OPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 34.79 | 53.06 | |
| SeqKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 26.11 | 42.87 |