List your company

Jailbreak Attack on Advbench subset

96ASRCodeAttack
-3.8422.084873.92May 15, 2024Sep 15, 2024Jan 16, 2025May 19, 2025Sep 19, 2025Jan 20, 2026May 23, 2026

Evaluation results

MethodLinks
CodeAttackTarget Model=DeepSeek-R12026.0596———————————————77.2
AE-CoTTarget Model=DeepSeek-R12026.0596———————————————78.8
AE-CoTTarget Model=Qwen32026.0596———————————————80
AE-CoTTarget Model=Gemini-2.52026.0596———————————————82.8
CodeAttackTarget Model=Qwen32026.0594———————————————76.4
ICRTTarget Model=Gemini-2.52026.0594———————————————80
H-CoTTarget Model=DeepSeek-R12026.0592———————————————74.4
H-CoTTarget Model=Qwen32026.0592———————————————74.8
AE-CoTTarget Model=o1-mini2026.0592———————————————70.4
ADCTarget Model=GPT-3.52024.0590————————————————
H-CoTTarget Model=Gemini-2.52026.0590———————————————66.4
AE-CoTTarget Model=o3-mini2026.0588———————————————72
GCGTarget Model=GPT-3.52024.0587————————————————
ICRTTarget Model=DeepSeek-R12026.0586———————————————77.2
H-CoTTarget Model=o3-mini2026.0586———————————————70.4
PAPTarget Model=DeepSeek-R12026.0584———————————————71.2
FlipAttackAttack Type=Black-box2024.1078.12———————81.9280.9665.1978.8483.8586.9284.4262.88—
ICRTTarget Model=Qwen32026.0578———————————————76.4
CL-GSOTarget Model=DeepSeek-R12026.0574———————————————59.2
ICRTTarget Model=o1-mini2026.0574———————————————61.2
ReNeLLMAttack Type=Black-box2024.1066.18———————94.0488.2789.6270.7783.0827.129.2367.31—
ArtPromptAttack Type=Black-box2024.1063.52———————93.7568.6584.8178.0683.462516.7357.69—
PAIRTarget Model=GPT-42024.0562————————————————
CL-GSOTarget Model=o1-mini2026.0562———————————————49.6
DeepInceptionAttack Type=Black-box2024.1061.55———————75.0579.1780.4666.1569.0418.0815.9688.46—
PAIRTarget Model=GPT-3.52024.0560————————————————
PAPTarget Model=Gemini-2.52026.0558———————————————48.8
CodeChameleonAttack Type=Black-box2024.1056.2———————85.5896.3584.4223.8562.3137.120.7759.23—
CodeAttackTarget Model=o3-mini2026.0556———————————————44.8
CodeAttackTarget Model=o1-mini2026.0554———————————————43.2
H-CoTTarget Model=o1-mini2026.0554———————————————60
AE-CoTTarget Model=GPT-52026.0554———————————————71.6
ADCTarget Model=GPT-42024.0552————————————————
CL-GSOTarget Model=Gemini-2.52026.0552———————————————42.4
GCGTarget Model=GPT-42024.0547————————————————
DRAAttack Type=Black-box2024.1044.26———————94.6278.8577.319508.2700—
CL-GSOTarget Model=Qwen32026.0536———————————————28.8
H-CoTTarget Model=GPT-52026.0536———————————————29.6
AutoDANAttack Type=White-box2024.1033.39———————72.3124.0436.9219.2327.1225.585.7756.15—
TAPAttack Type=Black-box2024.1033.28———————72.5356.654.6745.179.2312.691.1514.23—
GPTFUZZERAttack Type=Black-box2024.1033.24———————40.548.8544.0436.9234.6220040.96—
CL-GSOTarget Model=o3-mini2026.0532———————————————25.6
Base64Attack Type=Black-box2024.1031.73———————71.350.3882.691.3513.080.190.5884.23—
PAIRAttack Type=Black-box2024.1030.65———————71.5445.7447.4433.5312.515.389.629.42—
CL-GSOTarget Model=GPT-52026.0530———————————————24
PromptAttackAttack Type=Black-box2024.1028.85———————37.6926.1528.2723.2732.8822.8832.527.12—
CodeAttackTarget Model=Gemini-2.52026.0524———————————————19.2
PAPTarget Model=GPT-52026.0518———————————————14.4
ArtPromptTarget Model=Qwen32026.0516———————————————12.8
ICRTTarget Model=GPT-52026.0516———————————————12.8
ArtPromptTarget Model=DeepSeek-R12026.0514———————————————11.2
SelfCipherAttack Type=Black-box2024.1011.88———————0.5800.1959.6225.776.7302.12—
ArtPromptTarget Model=o3-mini2026.0510———————————————8
ICRTTarget Model=o3-mini2026.0510———————————————8
GCGAttack Type=White-box2024.107.5———————301.731.351.543.4613.460.967.5—
ArtPromptTarget Model=o1-mini2026.056———————————————4.8
PAPTarget Model=o1-mini2026.056———————————————4.8
COLD-AttackAttack Type=White-box2024.105.72———————18.652.121.351.735.5811.542.122.69—
MACAttack Type=White-box2024.104.93———————18.080.380.580.962.511.731.154.04—
PAPTarget Model=Qwen32026.054———————————————3.2
CodeAttackTarget Model=GPT-52026.052———————————————1.6
ArtPromptTarget Model=GPT-52026.050———————————————0
ArtPromptTarget Model=Gemini-2.52026.050———————————————0
PAPTarget Model=o3-mini2026.050———————————————0
DACAT2I prompt format adaptation=true2026.04—32302826363840—————————
Direct AttackT2I prompt format adaptation=true2026.04—30262426343234—————————
ETCHLLM Engine=Qwen3-Max, VLM Engine=Qwen3-VL-Plus, Maximum number of budget (T)=22026.04—62704442826276—————————
JailBrokenT2I prompt format adaptation=true2026.04—36382426404242—————————
MultiLingualT2I prompt format adaptation=true2026.04—30282624323634—————————
ReNeLLMT2I prompt format adaptation=true2026.04—34343028384038—————————
SneakyPromptT2I prompt format adaptation=true2026.04—30303028363836—————————
Template AttackT2I prompt format adaptation=true2026.04—30282824323638—————————