Text-based Jailbreak on AdvBench-M OOD
61.3ASR (OOD)CIDER
Evaluation Results
| Method | Links | |
|---|---|---|
| CIDERBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 61.3 | |
| CIDERBase Model=LLaVA-1.5-13B2026.04 | 61.3 | |
| SelfReminderBase Model=LLaVA-1.5-13B2026.04 | 42.65 | |
| JailGuardBase Model=LLaVA-1.5-13B2026.04 | 40.02 | |
| MirrorCheckBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 30.15 | |
| MirrorCheckBase Model=LLaVA-1.5-13B2026.04 | 30.15 | |
| JailGuardBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 27.49 | |
| ECSOBase Model=LLaVA-1.5-13B2026.04 | 22.09 | |
| ECSOBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 17.06 | |
| ASTRABase Model=LLaVA-1.5-13B2026.04 | 13.48 | |
| VLMGuardBase Model=LLaVA-1.5-13B2026.04 | 9.84 | |
| VLMGuardBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=Internal Defense2026.04 | 8.66 | |
| SelfReminderBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 6.72 | |
| ASTRABase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=Internal Defense2026.04 | 6.33 | |
| VLMShieldBase VLM=Qwen2.5-VL-7B-Instruct, Defense Category=External Defense2026.04 | 0.41 | |
| VLMShieldBase Model=LLaVA-1.5-13B2026.04 | 0.41 |