Safety Evaluation on UnsafeBench (test)
81F1 ScorePerspectiveVision
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PerspectiveVisionopensourced=false2025.01 | 81 | — | |
| ImageGuard2025.01 | 68.3 | 77.7 | |
| GPT-4o2025.01 | 55.5 | — | |
| LlavaGuard2025.01 | 53.7 | — | |
| Claude3.5-sonnet2025.01 | 53.4 | — | |
| Idefics22025.01 | 53 | — | |
| Multiheadedopensourced=false2025.01 | 50 | — | |
| InternVL22025.01 | 43.4 | — | |
| Gemini1.5-pro2025.01 | 35.8 | — | |
| SD_filter2025.01 | 32 | — | |
| LlavaNext2025.01 | 26.4 | — |