Image-based Jailbreak on HADES OOD
80.02Attack Success Rate (ASR)SelfReminder
Evaluation Results
| Method | Links | |
|---|---|---|
| SelfReminderBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 80.02 | |
| SelfReminderBase Model=LLaVA-1.5-13B2026.04 | 75.32 | |
| CIDERBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 51.86 | |
| CIDERBase Model=LLaVA-1.5-13B2026.04 | 51.86 | |
| JailGuardBase Model=LLaVA-1.5-13B2026.04 | 38.33 | |
| ECSOBase Model=LLaVA-1.5-13B2026.04 | 31.32 | |
| ECSOBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 30.74 | |
| MirrorCheckBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 23.09 | |
| MirrorCheckBase Model=LLaVA-1.5-13B2026.04 | 23.09 | |
| VLMGuardBase Model=LLaVA-1.5-13B2026.04 | 22.95 | |
| JailGuardBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 21.04 | |
| ASTRABase Model=LLaVA-1.5-13B2026.04 | 14.86 | |
| VLMGuardBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=Internal Defense2026.04 | 13.96 | |
| ASTRABase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=Internal Defense2026.04 | 7.83 | |
| VLMShieldBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 2.13 | |
| VLMShieldBase Model=LLaVA-1.5-13B2026.04 | 2.13 |