Science and general knowledge on OOD Benchmarks Aggregate
62.5Average OOD ScoreSRFT [13]
Evaluation Results
| Method | Links | |
|---|---|---|
| SRFT [13]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 62.5 | |
| SFT→RLBase Model=Qwen2.5-Math-7B, Source=verl2026.04 | 59.9 | |
| ReLIFT [12]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 59.5 | |
| HPT [15]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 59 | |
| Prefix-RFT [14]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 58.4 | |
| LUFFY [11]Base Model=Qwen2.5-Math-7B, Source=reproduction attempt2026.04 | 58 | |
| LUFFY [11]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 57.8 | |
| ReLIFT [12]Base Model=Qwen2.5-Math-7B, Source=reproduction attempt2026.04 | 56.2 | |
| SFTBase Model=Qwen2.5-Math-7B, Source=verl2026.04 | 52.4 |