Malicious Prompt Detection on JailbreakV_28K Image-based (test)
0.19FNRVLMShield
Evaluation Results
| Method | Links | |
|---|---|---|
| VLMShieldTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 0.19 | |
| VLMShieldTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 0.19 | |
| ASTRATarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 2.14 | |
| ASTRATarget VLM Architecture=LLaVA-1.5-13B2026.04 | 5.21 | |
| VLMGuardTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 11.82 | |
| JailGuardTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 14 | |
| VLMGuardTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 16.37 | |
| MirrorCheckTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 17.19 | |
| MirrorCheckTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 17.19 | |
| JailGuardTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 22.05 | |
| SelfReminderTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 34.8 | |
| CIDERTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 37.2 | |
| CIDERTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 37.2 | |
| ECSOTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 39.68 | |
| ECSOTarget VLM Architecture=Qwen2.5-VL-7B-Instruct2026.04 | 43.06 | |
| SelfReminderTarget VLM Architecture=LLaVA-1.5-13B2026.04 | 80.04 |