Visual Multi-Choice on PoPe
88.3AccuracyRejection-sampling
Evaluation results
| Method | Links | |
|---|---|---|
| Rejection-samplingAlignment method=Rejection-sampling, Base Model=LLaVA-1.5-13b2024.02 | 88.3 | |
| SteerLMAlignment method=SteerLM, Base Model=LLaVA-1.5-13b2024.02 | 87.8 | |
| DPOAlignment method=DPO, Base Model=LLaVA-1.5-13b2024.02 | 87 | |
| LLAVA-RLHF-13bModel=LLAVA-RLHF-13b2024.02 | 86.9 | |
| LLaVA-1.5-13bModel=LLaVA-1.5-13b2024.02 | 85.9 | |
| Standard SFTAlignment method=Standard SFT, Base Model=LLaVA-1.5-13b2024.02 | 85 |