Safety Evaluation on HADES-Dataset
43.67HADES ScoreQwen-VL-Max
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen-VL-MaxEvaluator=Llama-Guard-3-8B, System Prompt Setting=None2024.11 | 43.67 | 96.22 | 96.51 | 97.23 | 95.49 | |
| GPT-4oEvaluator=Llama-Guard-3-8B, System Prompt Setting=None2024.11 | 2.62 | 97.96 | 97.82 | 97.09 | 97.53 | |
| GPT-4o-MiniEvaluator=Llama-Guard-3-8B, System Prompt Setting=None2024.11 | 2.62 | 97.82 | 97.23 | 96.65 | 94.76 | |
| Claude-3.5-SonnetEvaluator=Llama-Guard-3-8B, System Prompt Setting=None2024.11 | 0 | 35.66 | 42.07 | 29.84 | 9.46 |