Mathematical Reasoning on BigMath (test)
48.4True AccuracyNon-hacking
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Non-hackingBase Model=Qwen2.5-3B-Instruct2026.04 | 48.4 | — | |
| RFT+GRIFTBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=102026.04 | 37.1 | 45.7 | |
| RFT+TraceBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=102026.04 | 35 | 47.1 | |
| RFT+RandomBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=102026.04 | 32 | 45.8 | |
| Starting-pointBase Model=Qwen2.5-3B-Instruct, Checkpoint Step=102026.04 | 30.5 | 42 | |
| No-interventionBase Model=Qwen2.5-3B-Instruct2026.04 | 5.7 | 67.9 |