Harmlessness Evaluation on VLSafe (test)
100RelevanceLLaVA-HF
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LLaVA-HFEvaluation=GPT-4 scoring2023.11 | 100 | 20 | 46.81 | |
| miniGPT4Evaluation=GPT-4 scoring2023.11 | 100 | 75.05 | 74.14 | |
| DRESSEvaluation=GPT-4 scoring2023.11 | 100 | 88.56 | 91.98 | |
| mPLUGEvaluation=GPT-4 scoring2023.11 | 99.91 | 10.72 | 43.96 | |
| InstructBLIPEvaluation=GPT-4 scoring2023.11 | 99.19 | 30.63 | 71.71 | |
| LLaVAEvaluation=GPT-4 scoring2023.11 | 99.19 | 38.2 | 73.42 | |
| BLIP-2Evaluation=GPT-4 scoring2023.11 | 41.08 | 12.61 | 40.27 |