Code Generation on APPS (fixed 500-problem slice)
79.4Pass@4CPPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CPPOBackbone=Qwen3.5-4B, K=42026.05 | 79.4 | 4,728 | 1.679 | |
| PlanSearchBackbone=Qwen3.5-4B, K=42026.05 | 55.4 | 5,535 | 1.001 | |
| Plan-and-SolveBackbone=Qwen3.5-4B, K=42026.05 | 53 | 7,326 | 0.723 | |
| Direct SolveBackbone=Qwen3.5-4B, K=42026.05 | 51.5 | 6,087 | 0.846 |