Mathematical Reasoning on OlympiadBench (Avg@16, Pass@16)
36.36Avg@16PriFT-mass
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PriFT-massBackbone=Qwen3-Base (8B)2026.06 | 36.36 | 64 | |
| PriFT-probBackbone=Qwen2.5-Math (7B)2026.06 | 36.14 | 65.78 | |
| PriFT-massBackbone=Qwen2.5-Math (7B)2026.06 | 35.75 | 61.78 | |
| TALRBackbone=Qwen2.5-Math (7B)2026.06 | 34.71 | 59.85 | |
| DFTBackbone=Qwen2.5-Math (7B)2026.06 | 34.5 | 56 | |
| PriFT-probBackbone=Qwen3-Base (8B)2026.06 | 34.2 | 63.85 | |
| ASFTBackbone=Qwen2.5-Math (7B)2026.06 | 33.95 | 63.11 | |
| DFTBackbone=Qwen3-Base (8B)2026.06 | 33.34 | 58.07 | |
| IDFTBackbone=Qwen3-Base (8B)2026.06 | 33.2 | 49.33 | |
| ASFTBackbone=Qwen3-Base (8B)2026.06 | 32.41 | 63.41 | |
| IDFTBackbone=Qwen2.5-Math (7B)2026.06 | 31.71 | 49.78 | |
| TALRBackbone=Qwen3-Base (8B)2026.06 | 31.7 | 58.67 | |
| OriginalBackbone=Qwen3-Base (8B)2026.06 | 27.23 | 66.07 | |
| SFTBackbone=Qwen3-Base (8B)2026.06 | 19 | 54.07 | |
| EAFTBackbone=Qwen3-Base (8B)2026.06 | 18.9 | 53.93 | |
| EAFTBackbone=Qwen2.5-Math (7B)2026.06 | 18.85 | 54.67 | |
| SFTBackbone=Qwen2.5-Math (7B)2026.06 | 18.79 | 54.52 | |
| OriginalBackbone=Qwen2.5-Math (7B)2026.06 | 17.66 | 58.67 |