Python Code Generation on MBPP+ (Pass@1)
46.8Pass@1Dynamic
Evaluation Results
| Method | Links | |
|---|---|---|
| DynamicModel=Qwen2.5-Coder-1.5B-Instruct, Fine-tuning method=dynamic constraints, Evaluation framework=bigcode-evaluation-harness2026.03 | 46.8 | |
| DAPOModel=Qwen2.5-Coder-1.5B-Instruct, Fine-tuning method=DAPO, Evaluation framework=bigcode-evaluation-harness2026.03 | 43.9 | |
| Qwen2.5-Coder-1.5B-InstructEvaluation framework=bigcode-evaluation-harness2026.03 | 40.2 |