General Knowledge Reasoning on MMLU (5-shot)
80.515-shot AccuracyBase
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BaseModel=Qwen3-14B2026.07 | 80.51 | 67.63 | |
| LP-SFTModel=Qwen3-14B2026.07 | 79.65 | — | |
| LP-SFTModel=Qwen3-14B2026.07 | 78.65 | 64.91 | |
| DFTModel=Qwen3-14B2026.07 | 78.32 | — | |
| GEMModel=Qwen3-14B2026.07 | 78.31 | — | |
| EAFTModel=Qwen3-14B2026.07 | 78.17 | — | |
| CEModel=Qwen3-14B2026.07 | 78.13 | — | |
| ASFTModel=Qwen3-14B2026.07 | 78.1 | 60.22 | |
| ASFTModel=Qwen3-14B2026.07 | 77.95 | — | |
| GEMModel=Qwen3-14B2026.07 | 77.08 | 58.1 | |
| EAFTModel=Qwen3-14B2026.07 | 76.8 | 60.28 | |
| CEModel=Qwen3-14B2026.07 | 76.78 | 60.13 | |
| DFTModel=Qwen3-14B2026.07 | 75.94 | 53.42 | |
| DFTModel=Qwen3-4B2026.07 | 69.63 | — | |
| LP-SFTModel=Qwen3-4B2026.07 | 69.55 | — | |
| ASFTModel=Qwen3-4B2026.07 | 69.54 | — | |
| CEModel=Qwen3-4B2026.07 | 68.93 | — | |
| GEMModel=Qwen3-4B2026.07 | 68.72 | — | |
| EAFTModel=Qwen3-4B2026.07 | 68.69 | — | |
| LP-SFTModel=Llama3.1-8B2026.07 | 62.11 | — | |
| CEModel=Llama3.1-8B2026.07 | 61.69 | — | |
| GEMModel=Llama3.1-8B2026.07 | 61.27 | — | |
| ASFTModel=Llama3.1-8B2026.07 | 60.77 | — | |
| EAFTModel=Llama3.1-8B2026.07 | 60.6 | — | |
| DFTModel=Llama3.1-8B2026.07 | 59.44 | — |