List your company

Jailbreaking on JailbreakBench (ASR via Detoxify & Llama Guard 3)

0ASR (Detoxify)VAE-JB
-0.442.535.58.47Apr 11, 2026

Evaluation results

MethodLinks
VAE-JBModel=InternVL, epsilon (ε)=322026.0402
VAE-JBModel=MiniGPT-4, epsilon (ε)=32, transfer=true2026.04072
BAPModel=Qwen-VL, epsilon (ε)=322026.0406
BAPModel=InternVL, epsilon (ε)=322026.0401
BAPModel=MiniGPT-4, epsilon (ε)=32, transfer=true2026.04073
JB-PcsModel=Qwen-VL, epsilon (ε)=–2026.0403
JB-PcsModel=InternVL, epsilon (ε)=–2026.0400
Attention-Guided Visual JailbreakingModel=MiniGPT-4, epsilon (ε)=16, transfer=true2026.04073
Attention-Guided Visual JailbreakingModel=MiniGPT-4, epsilon (ε)=32, transfer=true2026.04079
BAPModel=LLaVA-1.5, epsilon (ε)=322026.04164
VAE-JBModel=LLaVA-1.5, epsilon (ε)=322026.04269
JB-PcsModel=LLaVA-1.5, epsilon (ε)=–2026.04272
JB-PcsModel=MiniGPT-4, epsilon (ε)=–, transfer=true2026.04262
Attention-Guided Visual JailbreakingModel=InternVL, epsilon (ε)=322026.0422
Attention-Guided Visual JailbreakingModel=InternVL, epsilon (ε)=162026.04319
Attention-Guided Visual JailbreakingModel=LLaVA-1.5, epsilon (ε)=162026.04464
Attention-Guided Visual JailbreakingModel=Qwen-VL, epsilon (ε)=162026.04545
Attention-Guided Visual JailbreakingModel=LLaVA-1.5, epsilon (ε)=322026.046.184
VAE-JBModel=Qwen-VL, epsilon (ε)=322026.04761
Attention-Guided Visual JailbreakingModel=Qwen-VL, epsilon (ε)=322026.041192