Coding on MBPP (test)
74.3Pass@1Base
Evaluation Results
| Method | Links | |
|---|---|---|
| BaseBackbone=Qwen2.5-3B-Instruct2026.05 | 74.3 | |
| Low-SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 74.1 | |
| DFTBackbone=Qwen2.5-3B-Instruct2026.05 | 72.2 | |
| Iter-SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 71.7 | |
| STMBackbone=Qwen2.5-3B-Instruct2026.05 | 70.9 | |
| Anchored LearningBackbone=Qwen2.5-3B-Instruct2026.05 | 70.6 | |
| Self-SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 70.1 | |
| SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 69.6 | |
| KL-SFTBackbone=Qwen2.5-3B-Instruct2026.05 | 68 |