List your company

Adversarial Attack Transfer on 388 harmful behaviors

81.19Attack Success Rate (ASR)I-GCG
-3.247618.673740.59562.5163Jun 4, 2026

Evaluation results

MethodLinks
I-GCGTarget Model=Vicuna 7B 1.5v2026.0681.19
Universal SlotGCGBase Attack=I-GCG, Target Model=Vicuna 7B 1.5v2026.0679.9
Universal SlotGCGBase Attack=AttnGCG, Target Model=Vicuna 7B 1.5v2026.0676.8
GCGTarget Model=Vicuna 7B 1.5v2026.0670.1
AttnGCGTarget Model=Vicuna 7B 1.5v2026.0664.43
GCG-HijTarget Model=Vicuna 7B 1.5v2026.0663.66
Universal SlotGCGBase Attack=GCG, Target Model=Vicuna 7B 1.5v2026.0663.4
Universal SlotGCGBase Attack=GCG-Hij, Target Model=Vicuna 7B 1.5v2026.0661.34
Universal SlotGCGBase Attack=GCG, Target Model=GPT-3.5-turbo2026.0650.77
Universal SlotGCGBase Attack=AttnGCG, Target Model=GPT-3.5-turbo2026.0643.04
Universal SlotGCGBase Attack=GCG-Hij, Target Model=GPT-3.5-turbo2026.0641.49
Universal SlotGCGBase Attack=I-GCG, Target Model=GPT-3.5-turbo2026.0640.46
GCG-HijTarget Model=GPT-3.5-turbo2026.0637.89
AttnGCGTarget Model=GPT-3.5-turbo2026.0625.52
Universal SlotGCGBase Attack=I-GCG, Target Model=Average2026.0624.83
Universal SlotGCGBase Attack=GCG, Target Model=Average2026.0624.73
Universal SlotGCGBase Attack=AttnGCG, Target Model=Average2026.0624.68
Universal SlotGCGBase Attack=GCG-Hij, Target Model=Average2026.0622.38
GCG-HijTarget Model=Average2026.0620.62
I-GCGTarget Model=Average2026.0618.92
AttnGCGTarget Model=Average2026.0618.2
GCGTarget Model=Average2026.0614.9
I-GCGTarget Model=GPT-3.5-turbo2026.0612.89
Universal SlotGCGBase Attack=GCG-Hij, Target Model=Gemini 2.5 Pro2026.066.7
Universal SlotGCGBase Attack=GCG-Hij, Target Model=Gemini 2.0 Flash2026.064.12
Universal SlotGCGBase Attack=GCG, Target Model=Gemini 2.5 Pro2026.063.61
GCGTarget Model=GPT-3.5-turbo2026.063.09
Universal SlotGCGBase Attack=GCG, Target Model=Gemini 2.0 Flash2026.062.06
Universal SlotGCGBase Attack=GCG, Target Model=GPT-4o2026.061.8
Universal SlotGCGBase Attack=AttnGCG, Target Model=GPT-4o2026.061.55
GCG-HijTarget Model=Gemini 2.5 Pro2026.061.55
GCGTarget Model=Gemini 2.0 Flash2026.061.29
Universal SlotGCGBase Attack=I-GCG, Target Model=GPT-4o2026.060.77
Universal SlotGCGBase Attack=I-GCG, Target Model=Gemini 2.0 Flash2026.060.77
AttnGCGTarget Model=GPT-4o2026.060.52
I-GCGTarget Model=Gemini 2.5 Pro2026.060.52
Universal SlotGCGBase Attack=GCG-Hij, Target Model=GPT-4o2026.060.26
AttnGCGTarget Model=Gemini 2.0 Flash2026.060.26
AttnGCGTarget Model=Gemini 2.5 Pro2026.060.26
Universal SlotGCGBase Attack=I-GCG, Target Model=Gemini 2.5 Pro2026.060.26
GCGTarget Model=GPT-4o2026.060
I-GCGTarget Model=GPT-4o2026.060
GCG-HijTarget Model=GPT-4o2026.060
Universal SlotGCGBase Attack=AttnGCG, Target Model=Gemini 2.0 Flash2026.060
I-GCGTarget Model=Gemini 2.0 Flash2026.060
GCG-HijTarget Model=Gemini 2.0 Flash2026.060
GCGTarget Model=Gemini 2.5 Pro2026.060
Universal SlotGCGBase Attack=AttnGCG, Target Model=Gemini 2.5 Pro2026.060