Code Generation on LiveCodeBench (OOD)
19.2AccuracyRL-PLUS
Evaluation Results
| Method | Links | |
|---|---|---|
| RL-PLUSBase model=Qwen2.5-Math-7B2025.07 | 19.2 | |
| GRPOBase model=Qwen2.5-Math-7B2025.07 | 15.3 | |
| Base ModelBase model=Qwen2.5-Math-7B2025.07 | 14.9 | |
| SFT+GRPOBase model=Qwen2.5-Math-7B2025.07 | 9.7 | |
| SFTBase model=Qwen2.5-Math-7B2025.07 | 8.1 |