List your company

Jailbreak Attack on AdvBench (ASR, MSS, ASS, LFS)

100ASREmbedding Attack (Standard)
15.7637.6359.581.37Feb 13, 2025

Evaluation results

MethodLinks
Embedding Attack (Standard)Attack Algorithm=Embedding Attack, Initialization=Standard, Target Model=Llama-22025.021001212.51.71
Embedding Attack (Baseline)Attack Algorithm=Embedding Attack, Initialization=Baseline, Target Model=Llama-22025.0210033.721.16
Embedding Attack (25-CRI)Attack Algorithm=Embedding Attack, Initialization=25-CRI, Target Model=Llama-22025.0210011.410.08
Embedding Attack (5-CRI)Attack Algorithm=Embedding Attack, Initialization=5-CRI, Target Model=Llama-22025.0210022.60.51
Embedding Attack (1-CRI)Attack Algorithm=Embedding Attack, Initialization=1-CRI, Target Model=Llama-22025.0210076.843.72
Embedding Attack (Baseline)Attack Algorithm=Embedding Attack, Initialization=Baseline, Target Model=Vicuna2025.0210033.960.95
Embedding Attack (5-CRI)Attack Algorithm=Embedding Attack, Initialization=5-CRI, Target Model=Vicuna2025.0210022.580.51
AutoDAN-GA (Standard)Attack Algorithm=AutoDAN-GA, Initialization=Standard, Target Model=Vicuna2025.0210022.210.84
AutoDAN-GA (25-CRI)Attack Algorithm=AutoDAN-GA, Initialization=25-CRI, Target Model=Vicuna2025.0210022.20.65
AutoDAN-HGA (Standard)Attack Algorithm=AutoDAN-HGA, Initialization=Standard, Target Model=Vicuna2025.0210022.320.84
AutoDAN-HGA (25-CRI)Attack Algorithm=AutoDAN-HGA, Initialization=25-CRI, Target Model=Vicuna2025.0210022.380.37
GCG (5-CRI)Attack Algorithm=GCG, Initialization=5-CRI, Target Model=Llama-32025.02100111.290.63
GCG (1-CRI)Attack Algorithm=GCG, Initialization=1-CRI, Target Model=Llama-32025.02100115.121
Embedding Attack (25-CRI)Attack Algorithm=Embedding Attack, Initialization=25-CRI, Target Model=Llama-32025.0210078.564.36
Embedding Attack (5-CRI)Attack Algorithm=Embedding Attack, Initialization=5-CRI, Target Model=Llama-32025.0210022.440.6
Embedding Attack (1-CRI)Attack Algorithm=Embedding Attack, Initialization=1-CRI, Target Model=Llama-32025.0210066.025.77
AutoDAN-GA (Standard)Attack Algorithm=AutoDAN-GA, Initialization=Standard, Target Model=Llama-32025.0210035.671.58
AutoDAN-GA (25-CRI)Attack Algorithm=AutoDAN-GA, Initialization=25-CRI, Target Model=Llama-32025.0210033.981.17
AutoDAN-HGA (Standard)Attack Algorithm=AutoDAN-HGA, Initialization=Standard, Target Model=Llama-32025.0210027.481.58
AutoDAN-HGA (25-CRI)Attack Algorithm=AutoDAN-HGA, Initialization=25-CRI, Target Model=Llama-32025.0210025.941.03
GCG (1-CRI)Attack Algorithm=GCG, Initialization=1-CRI, Target Model=Llama-22025.029911.680.33
GCG (Random)Attack Algorithm=GCG, Initialization=Random, Target Model=Vicuna2025.0298915.80.58
GCG (1-CRI)Attack Algorithm=GCG, Initialization=1-CRI, Target Model=Vicuna2025.029811.010.25
GCG (25-CRI)Attack Algorithm=GCG, Initialization=25-CRI, Target Model=Llama-32025.0298418.40.74
Embedding Attack (Standard)Attack Algorithm=Embedding Attack, Initialization=Standard, Target Model=Llama-32025.029822283.3
Embedding Attack (Baseline)Attack Algorithm=Embedding Attack, Initialization=Baseline, Target Model=Llama-32025.029846.362.12
GCG (25-CRI)Attack Algorithm=GCG, Initialization=25-CRI, Target Model=Llama-22025.029712.040.21
GCG (5-CRI)Attack Algorithm=GCG, Initialization=5-CRI, Target Model=Llama-22025.029711.20.28
GCG (Standard)Attack Algorithm=GCG, Initialization=Standard, Target Model=Vicuna2025.0297815.750.8
GCG (25-CRI)Attack Algorithm=GCG, Initialization=25-CRI, Target Model=Vicuna2025.029712.830.23
GCG (5-CRI)Attack Algorithm=GCG, Initialization=5-CRI, Target Model=Vicuna2025.029711.610.31
AutoDAN-HGA (25-CRI)Attack Algorithm=AutoDAN-HGA, Initialization=25-CRI, Target Model=Llama-22025.02924.513.81.48
GCG (Standard)Attack Algorithm=GCG, Initialization=Standard, Target Model=Llama-22025.029064106.512.29
GCG (Standard)Attack Algorithm=GCG, Initialization=Standard, Target Model=Llama-32025.028976118.271.84
Embedding Attack (Standard)Attack Algorithm=Embedding Attack, Initialization=Standard, Target Model=Vicuna2025.02862736.382.99
Embedding Attack (25-CRI)Attack Algorithm=Embedding Attack, Initialization=25-CRI, Target Model=Vicuna2025.0286921.582.29
Embedding Attack (1-CRI)Attack Algorithm=Embedding Attack, Initialization=1-CRI, Target Model=Vicuna2025.0286921.263.63
GCG (Random)Attack Algorithm=GCG, Initialization=Random, Target Model=Llama-32025.027889132.261.76
GCG (Random)Attack Algorithm=GCG, Initialization=Random, Target Model=Llama-22025.027680114.931.9
AutoDAN-HGA (Standard)Attack Algorithm=AutoDAN-HGA, Initialization=Standard, Target Model=Llama-22025.026719.529.62.25
AutoDAN-GA (25-CRI)Attack Algorithm=AutoDAN-GA, Initialization=25-CRI, Target Model=Llama-22025.023089.551.9
AutoDAN-GA (Standard)Attack Algorithm=AutoDAN-GA, Initialization=Standard, Target Model=Llama-22025.021999.842.25