Coding on MBPP+ (pass@1, Overall Avg.)
54.8Pass@1Anchored Learning
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Anchored LearningBackbone=Llama-3.2-3B-Instruct2026.05 | 54.8 | 52 | |
| Low-SFTBackbone=Llama-3.2-3B-Instruct2026.05 | 54 | 45.6 | |
| STMBackbone=Llama-3.2-3B-Instruct2026.05 | 53.4 | 41.8 | |
| BaseBackbone=Llama-3.2-3B-Instruct2026.05 | 51.6 | 39.4 | |
| Iter-SFTBackbone=Llama-3.2-3B-Instruct2026.05 | 49.4 | 39.6 | |
| Self-SFTBackbone=Llama-3.2-3B-Instruct2026.05 | 46.6 | 37.8 | |
| DFTBackbone=Llama-3.2-3B-Instruct2026.05 | 22 | 24.2 | |
| KL-SFTBackbone=Llama-3.2-3B-Instruct2026.05 | 21.7 | 20.7 | |
| SFTBackbone=Llama-3.2-3B-Instruct2026.05 | 1.2 | 19.1 |