List your company

Jailbreaking on AdvBench (BERT Score & Harmful Scores)

4.84BERT ScoreJULI
0.44081.58292.7253.8671May 17, 2025

Evaluation results

MethodLinks
JULIModel=Qwen2.5-1.5B-INST2025.054.844.763.73
JULIModel=Llama3-3B-INST2025.054.814.663.68
FLIPAPI Compatibility=true2025.054.762.060.5
JULITarget Model=Llama2-7B-CHAT, Defense=DEEPALIGN, Evaluation Setting=open-source setting2025.054.754.723.98
EDAPI Compatibility=false2025.054.434.283.36
GCGAPI Compatibility=false2025.054.421.930.48
EDModel=Llama3-8B-INST2025.054.394.13.02
JULIModel=Gemini-2.5-Pro2025.054.374.193.19
JULIModel=Llama3-8B-INST2025.054.334.573.44
LINTAPI Compatibility=false2025.054.051.950.77
EDModel=Llama3-3B-INST2025.0544.222.99
NaiveAPI Compatibility=true2025.053.961.540.45
FLIPModel=Llama3-3B-INST2025.053.952.80.81
JULIModel=Llama2-7B-CHAT2025.053.944.223.5
JULIAPI Compatibility=false2025.053.73.082.35
EDModel=Llama2-7B-CHAT2025.053.692.961.84
LINTModel=Qwen2.5-1.5B-INST2025.053.654.213.13
NaiveModel=Qwen2.5-1.5B-INST2025.053.632.792.02
FLIPModel=Llama3-8B-INST2025.053.542.870.87
FLIPModel=Llama3-8B-INST2025.053.542.870.87
WTSModel=Qwen2.5-1.5B-INST2025.053.543.822.62
LINTModel=Llama2-7B-CHAT2025.053.423.72.22
FLIPModel=Gemini-2.5-Flash2025.053.413.131.33
FLIPModel=Qwen2.5-1.5B-INST2025.053.342.440.79
EDModel=Qwen2.5-1.5B-INST2025.053.241.260.6
WTSAPI Compatibility=false2025.053.141.370.42
GCGModel=Qwen2.5-1.5B-INST2025.053.132.772.05
BaseAPI Compatibility=true2025.053.071.40.41
BaseModel=Qwen2.5-1.5B-INST2025.052.983.042.14
JULI (API)API Compatibility=true2025.052.951.920.75
JULIModel=Llama3-8B-INST2025.052.913.122.21
LINTModel=Llama3-3B-INST2025.052.653.682.16
LINTModel=Llama3-8B-INST2025.052.633.772.25
JULIModel=Gemini-2.5-Flash2025.052.582.521.74
WTSModel=Llama3-8B-INST2025.052.462.381.26
NaiveModel=Llama2-7B-CHAT2025.052.412.511.74
FLIPModel=Gemini-2.5-Pro2025.052.392.521.38
GCGModel=Llama3-3B-INST2025.052.281.710.76
NaiveModel=Gemini-2.5-Flash2025.052.092.091.29
NaiveModel=Llama3-8B-INST2025.052.071.480.41
NaiveModel=Llama3-8B-INST2025.052.071.480.41
NaiveModel=Llama3-3B-INST2025.051.981.210.17
FLIPModel=Llama2-7B-CHAT2025.051.921.040.07
WTSModel=Llama2-7B-CHAT2025.051.871.640.56
GCGModel=Llama3-8B-INST2025.051.821.380.35
WTSModel=Llama3-3B-INST2025.051.811.560.44
NaiveModel=Gemini-2.5-Pro2025.051.71.911.21
BaseModel=Llama3-8B-INST2025.051.641.40.39
BaseModel=Llama3-8B-INST2025.051.641.40.39
GCGModel=Llama2-7B-CHAT2025.051.561.40.44
BaseModel=Llama3-3B-INST2025.051.321.210.21
NaiveTarget Model=Llama2-7B-CHAT, Defense=DEEPALIGN, Evaluation Setting=open-source setting2025.051.081.010
BaseModel=Llama2-7B-CHAT2025.050.791.040.04
BaseModel=Gemini-2.5-Flash2025.050.6410.02
BaseModel=Gemini-2.5-Pro2025.050.611.020.06