Visual Instruction Following on LLaVA-Bench
79.1Overall ScoreDPO
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DPOAlignment method=DPO, Base Model=LLaVA-1.5-13b2024.02 | 79.1 | — | — | — | — | — | |
| LLAVA-RLHF-13bModel=LLAVA-RLHF-13b2024.02 | 76.8 | — | — | — | — | — | |
| LLaVA-1.5-13bModel=LLaVA-1.5-13b2024.02 | 73.1 | — | — | — | — | — | |
| Rejection-samplingAlignment method=Rejection-sampling, Base Model=LLaVA-1.5-13b2024.02 | 70.6 | — | — | — | — | — | |
| SteerLMAlignment method=SteerLM, Base Model=LLaVA-1.5-13b2024.02 | 67 | — | — | — | — | — | |
| Standard SFTAlignment method=Standard SFT, Base Model=LLaVA-1.5-13b2024.02 | 63.7 | — | — | — | — | — | |
| Fed-CMPFederated Learning Setting=M_T-T2026.03 | 43.2 | — | — | — | — | — | |
| FedAvg-TIEsFederated Learning Setting=M_T-T2026.03 | 41.5 | — | — | — | — | — | |
| FedAvg-DAREFederated Learning Setting=M_T-T2026.03 | 40.6 | — | — | — | — | — | |
| MOONFederated Learning Setting=M_T-T2026.03 | 40.1 | — | — | — | — | — | |
| FedAvg-TAFederated Learning Setting=M_T-T2026.03 | 40 | — | — | — | — | — | |
| FedAvgFederated Learning Setting=M_T-T2026.03 | 39.6 | — | — | — | — | — | |
| FedAdamFederated Learning Setting=M_T-T2026.03 | 39.2 | — | — | — | — | — | |
| Local TrainingFederated Learning Setting=M_T-T2026.03 | 38.5 | — | — | — | — | — | |
| FedProxFederated Learning Setting=M_T-T2026.03 | 36.9 | — | — | — | — | — | |
| DRESSBackbone=EVA-CLIP-Giant, LLM=Vicuna-13b-v1.52023.11 | — | — | — | — | — | 30.81 | |
| GPT-4V2023.11 | — | — | — | — | — | 38.42 | |
| InstructBLIP2023.11 | — | — | — | — | — | 27.32 | |
| LLaVAParameters=7B2023.09 | — | 75.1 | 75.4 | 92.3 | 81 | — | |
| LLaVAParameters=13B, Resolution=3362023.09 | — | 87.2 | 74.3 | 92.9 | 84.9 | — | |
| LLaVA2023.11 | — | — | — | — | — | 18.32 | |
| LLaVA-RLHFParameters=7B2023.09 | — | 93 | 79 | 109.5 | 94.1 | — | |
| LLaVA-RLHFParameters=13B, Resolution=3362023.09 | — | 93.9 | 82.5 | 110.1 | 95.6 | — | |
| LLaVA-SFT+Parameters=7B2023.09 | — | 88.8 | 74.6 | 95 | 86.3 | — | |
| LLaVA-SFT+Parameters=13B, Resolution=3362023.09 | — | 85.8 | 75.5 | 93.9 | 85.2 | — | |
| miniGPT42023.11 | — | — | — | — | — | 28.32 | |
| mPLUG2023.11 | — | — | — | — | — | 24.68 | |
| VIGCParameters=7B2023.09 | — | 83.3 | 80.6 | 93.1 | 85.8 | — | |
| VIGCParameters=13B, Resolution=3362023.09 | — | 88.9 | 77.4 | 93.5 | 86.8 | — |