Textual understanding on IFEval
71.2Strict AccuracyUni-DPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Uni-DPOModel=Gemma-2-9B-IT2025.06 | 71.2 | 72.8 | |
| Uni-DPO v0.2Model=Llama3-8B Instruct2025.06 | 66.9 | 73.2 | |
| SimPOModel=Llama3-8B Instruct2025.06 | 61.2 | 65.8 | |
| Uni-DPOModel=Llama3-8B Instruct2025.06 | 61.2 | 67.8 | |
| SimPO v0.2Model=Llama3-8B Instruct2025.06 | 60.8 | 68.6 | |
| SimPOModel=Gemma-2-9B-IT2025.06 | 60.4 | 67.7 | |
| SFTModel=Llama3-8B Instruct2025.06 | 59.4 | 62.2 | |
| SimPOModel=Qwen2.5-7B2025.06 | 43.6 | 48.2 | |
| Uni-DPOModel=Qwen2.5-7B2025.06 | 43.3 | 48.1 | |
| Uni-DPO v0.2Model=Llama3-8B Base2025.06 | 42.1 | 48.6 | |
| DPOModel=Llama3-8B Base2025.06 | 40.5 | 45.5 | |
| SimPOModel=Llama3-8B Base2025.06 | 40.5 | 45.7 | |
| SimPO v0.2Model=Llama3-8B Base2025.06 | 40.1 | 45.3 | |
| Uni-DPOModel=Llama3-8B Base2025.06 | 38.1 | 47.9 | |
| SFTModel=Llama3-8B Base2025.06 | 33.5 | 35.3 |