General Language Understanding on MMLU (5-shot Accuracy)
78.825-shot AccuracyLP-SFT
Evaluation Results
| Method | Links | |
|---|---|---|
| LP-SFTBackbone=Qwen3-14B, Few-shot prompt=5-shot2026.07 | 78.82 | |
| ASFTBackbone=Qwen3-14B, Few-shot prompt=5-shot2026.07 | 77.7 | |
| EAFTBackbone=Qwen3-14B, Few-shot prompt=5-shot2026.07 | 77.13 | |
| CEBackbone=Qwen3-14B, Few-shot prompt=5-shot2026.07 | 76.99 | |
| GEMBackbone=Qwen3-14B, Few-shot prompt=5-shot2026.07 | 76.96 | |
| DFTBackbone=Qwen3-14B, Few-shot prompt=5-shot2026.07 | 76.11 | |
| LP-SFTBackbone=Qwen3-4B, Few-shot prompt=5-shot2026.07 | 70.32 | |
| ASFTBackbone=Qwen3-4B, Few-shot prompt=5-shot2026.07 | 69.54 | |
| EAFTBackbone=Qwen3-4B, Few-shot prompt=5-shot2026.07 | 69.08 | |
| GEMBackbone=Qwen3-4B, Few-shot prompt=5-shot2026.07 | 68.99 | |
| CEBackbone=Qwen3-4B, Few-shot prompt=5-shot2026.07 | 68.9 | |
| DFTBackbone=Qwen3-4B, Few-shot prompt=5-shot2026.07 | 66.82 | |
| LP-SFTBackbone=Llama3.1-8B, Few-shot prompt=5-shot2026.07 | 62.48 | |
| EAFTBackbone=Llama3.1-8B, Few-shot prompt=5-shot2026.07 | 61.14 | |
| CEBackbone=Llama3.1-8B, Few-shot prompt=5-shot2026.07 | 61.11 | |
| GEMBackbone=Llama3.1-8B, Few-shot prompt=5-shot2026.07 | 61 | |
| ASFTBackbone=Llama3.1-8B, Few-shot prompt=5-shot2026.07 | 60.58 | |
| DFTBackbone=Llama3.1-8B, Few-shot prompt=5-shot2026.07 | 55.11 |