List your company

Jailbreak Attack on AdvBench (ASR, Avg.Score)

0Attack Success Rate (ASR)PAIR
-4235077Feb 26, 2026Mar 5, 2026Mar 12, 2026Mar 20, 2026Mar 27, 2026Apr 3, 2026Apr 11, 2026

Evaluation results

MethodLinks
PAIRTarget LLM=Gemini-2.5-flash2026.0200
PAIRTarget LLM=GPT-4o2026.0200
PAIRTarget LLM=Qwen32026.0200.02
TAPTarget LLM=Gemini-2.5-flash2026.0200
TAPTarget LLM=Claude-3.72026.0200
GPTFUZZERTarget LLM=Claude-3.72026.0200.04
PAIRTarget LLM=Claude-3.72026.0220.06
GPTFUZZERTarget LLM=Qwen32026.0220.06
TAPTarget LLM=Qwen32026.0240.12
TAPTarget LLM=Deepseek-Reasoner2026.0260.24
PAIRTarget LLM=Deepseek-Reasoner2026.0280.24
VAETarget Model=Claude-3.5, Source Model=Qwen-VL, Perturbation budget (epsilon)=32/2552026.048.5—
TAPTarget LLM=GPT-4o2026.02120.4
GPTFUZZERTarget LLM=GPT-4o2026.02120.4
Attention-Guided Visual JailbreakingTarget Model=Claude-3.5, Source Model=Qwen-VL, Perturbation budget (epsilon)=32/2552026.0412.1—
PAIRTarget LLM=Grok-32026.02140.52
VAETarget Model=GPT-4o, Source Model=Qwen-VL, Perturbation budget (epsilon)=32/2552026.0414.2—
VAETarget Model=Gemini-1.5, Source Model=Qwen-VL, Perturbation budget (epsilon)=32/2552026.0416.4—
Attention-Guided Visual JailbreakingTarget Model=GPT-4o, Source Model=Qwen-VL, Perturbation budget (epsilon)=32/2552026.0418.5—
GPTFUZZERTarget LLM=Deepseek-Reasoner2026.02241.04
GPTFUZZERTarget LLM=Gemini-2.5-flash2026.02281.22
Attention-Guided Visual JailbreakingTarget Model=Gemini-1.5, Source Model=Qwen-VL, Perturbation budget (epsilon)=32/2552026.0430.3—
CL-GSOTarget LLM=Claude-3.72026.02401.86
ICRTTarget LLM=Claude-3.72026.02401.6
CL-GSOTarget LLM=Grok-32026.02442.04
CL-GSOTarget LLM=Deepseek-Reasoner2026.02501.9
CL-GSOTarget LLM=Qwen32026.02502
GPTFUZZERTarget LLM=Grok-32026.02522.1
TAPTarget LLM=Grok-32026.02542.04
AutoDAN-Turbo-RTarget LLM=Gemini-2.5-flash2026.02702.52
AutoDAN-Turbo-RTarget LLM=Claude-3.72026.02742.88
ICRTTarget LLM=GPT-4o2026.02743.06
CL-GSOTarget LLM=GPT-4o2026.02782.64
CL-GSOTarget LLM=Gemini-2.5-flash2026.02802.6
AutoDAN-Turbo-RTarget LLM=Grok-32026.02843.1
ICRTTarget LLM=Qwen32026.02844
AutoDAN-Turbo-RTarget LLM=GPT-4o2026.02883.18
AutoDAN-Turbo-RTarget LLM=Deepseek-Reasoner2026.02883.36
AutoDAN-Turbo-RTarget LLM=Qwen32026.02883.32
ICRTTarget LLM=Deepseek-Reasoner2026.02884
ICRTTarget LLM=Gemini-2.5-flash2026.02924.52
ICRTTarget LLM=Grok-32026.02984.3
CC-BOSTarget LLM=Gemini-2.5-flash2026.021004.82
CC-BOSTarget LLM=Claude-3.72026.021003.14
CC-BOSTarget LLM=GPT-4o2026.021004.74
CC-BOSTarget LLM=Deepseek-Reasoner2026.021004.84
CC-BOSTarget LLM=Qwen32026.021004.88
CC-BOSTarget LLM=Grok-32026.021004.76