Multimodal Safety Evaluation on SafeBench
3.26FS ASRClaude-3.5-Sonnet
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Claude-3.5-SonnetEvaluator=Llama-Guard-3-8B, System Prompt=None2024.11 | 3.26 | 1.3 | 41.69 | 53.42 | 38.11 | 8.14 | |
| GPT-4oEvaluator=Llama-Guard-3-8B, System Prompt=None2024.11 | 6.19 | 3.91 | 96.42 | 97.39 | 96.42 | 96.74 | |
| GPT-4o-MiniEvaluator=Llama-Guard-3-8B, System Prompt=None2024.11 | 13.68 | 7.82 | 96.42 | 95.77 | 96.42 | 94.14 | |
| Qwen-VL-MaxEvaluator=Llama-Guard-3-8B, System Prompt=None2024.11 | 89.58 | 53.75 | 92.51 | 96.74 | 92.18 | 91.86 |