Code Generation on LiveCodeBench (Avg@16, Pass@16)
36.8LCB Avg@16PieceHint
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PieceHintBase model=DeepSeek-R1-Distill-Qwen-7B2026.04 | 36.8 | 55.3 | |
| GRPOBase model=DeepSeek-R1-Distill-Qwen-7B2026.04 | 31.9 | 51.6 | |
| DeepSeek-R1-Distill-Qwen-7BBase model=DeepSeek-R1-Distill-Qwen-7B2026.04 | 28.4 | 47.2 |