Mathematical Reasoning on ID Benchmarks Aggregate
57Average ID ScoreSFT→RL
Evaluation results
| Method | Links | |
|---|---|---|
| SFT→RLBase Model=Qwen2.5-Math-7B, Source=verl2026.04 | 57 | |
| SRFT [13]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 53.2 | |
| HPT [15]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 52.7 | |
| SFTBase Model=Qwen2.5-Math-7B, Source=verl2026.04 | 52.2 | |
| Prefix-RFT [14]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 51.8 | |
| LUFFY [11]Base Model=Qwen2.5-Math-7B, Source=reported2026.04 | 50.1 | |
| ReLIFT [12]Base Model=Qwen2.5-Math-7B, Source=reproduction attempt2026.04 | 48.8 | |
| LUFFY [11]Base Model=Qwen2.5-Math-7B, Source=reproduction attempt2026.04 | 46.3 |