Malicious Prompt Detection on JailBreakV GPT4V
99.05AUROCVLMGuard
Evaluation results
| Method | Links | |
|---|---|---|
| VLMGuardBackbone=Qwen2.5-VL-72B-Instruct2024.10 | 99.05 | |
| GuardRankBackbone=LLaVA-1.6-7B, Training Supervision=Fully supervised2024.10 | 98.03 | |
| VLMGuardBackbone=LLaVA-1.6-7B, Training Supervision=Unsupervised (no labeled data)2024.10 | 96.58 | |
| ASTRABackbone=Qwen2.5-VL-72B-Instruct2024.10 | 96.11 | |
| HiddenDetectBackbone=Qwen2.5-VL-72B-Instruct2024.10 | 85.71 |