List your company

Jailbreaking Attack Success on STRONGREJECT 40 held-out behaviors

94EVUSIHO
-3.7621.624772.38Jun 2, 2026

Evaluation results

MethodLinks
IHOTarget model=Qwen-7B2026.0694
IHOTarget model=Qwen-32B2026.0692
PAIR+Target model=Qwen-7B2026.0690
GCG+Target model=Qwen-7B2026.0689
JR1+Target model=Qwen-7B2026.0689
JR1+Target model=Qwen-32B2026.0688
PAIR+Target model=Qwen-32B2026.0687
GCG+Target model=Qwen-32B2026.0685
INP.Target model=Qwen-7B2026.0685
IHOTarget model=LLaMA-8B2026.0685
JR1+Target model=LLaMA-8B2026.0683
IHOTarget model=Qwen-7B+D2026.0682
IHOTarget model=CB2026.0682
BONTarget model=LLaMA-8B2026.0681
INP.Target model=Qwen-32B2026.0680
IHOTarget model=CB+D2026.0679
PAIRTarget model=Qwen-7B2026.0678
GCG+Target model=LLaMA-8B2026.0678
IHOTarget model=LAT2026.0678
IHOTarget model=CAT2026.0678
BONTarget model=Qwen-7B2026.0677
PAIR+Target model=LLaMA-8B2026.0676
PAIRTarget model=Qwen-32B2026.0675
BONTarget model=Qwen-32B2026.0675
INP.Target model=LLaMA-8B2026.0674
JR1+Target model=Qwen-7B+D2026.0672
JR1+Target model=CAT2026.0672
PAIR+Target model=Qwen-7B+D2026.0670
INP.Target model=Qwen-7B+D2026.0669
JR1+Target model=LAT2026.0667
PAIR+Target model=CAT2026.0667
PAIR+Target model=LAT2026.0665
INP.Target model=CB2026.0664
PAIR+Target model=CB2026.0662
INP.Target model=LAT2026.0662
BONTarget model=Qwen-7B+D2026.0661
INP.Target model=CAT2026.0658
INP.Target model=CB+D2026.0657
PAIRTarget model=LLaMA-8B2026.0656
JR1+Target model=CB2026.0654
Ample+Target model=Qwen-7B2026.0653
PAIR+Target model=CB+D2026.0652
BONTarget model=CB2026.0649
GCG+Target model=LAT2026.0648
JR1+Target model=CB+D2026.0644
PAIRTarget model=Qwen-7B+D2026.0643
PAIRTarget model=CAT2026.0641
PAIRTarget model=LAT2026.0638
BONTarget model=CB+D2026.0637
Ample+Target model=Qwen-32B2026.0636
PAIRTarget model=CB2026.0627
PAIRTarget model=CB+D2026.0621
GCG+Target model=CB2026.0617
Ample+Target model=LLaMA-8B2026.0613
Ample+Target model=Qwen-7B+D2026.069
Ample+Target model=CB2026.069
GCG+Target model=CAT2026.067
BONTarget model=LAT2026.063
Ample+Target model=CB+D2026.061
Ample+Target model=LAT2026.061
BONTarget model=CAT2026.061
Ample+Target model=CAT2026.060