List your company

Jailbreak Attack on Advbench Vicuna-33B Guard 100 prompts Original

0ASRNo Attack
-3.6420.9345.570.07Feb 24, 2024

Evaluation results

MethodLinks
No AttackLLM Model=Llama2-70B, Attack Version=Original2024.020
GCGLLM Model=Llama2-70B, Attack Version=Original2024.020
GCG + PRPLLM Model=Llama2-70B, Attack Version=PRP-W, Access Threat Model=white-box2024.021
GCG + PRPLLM Model=Llama2-70B, Attack Version=PRP-B, Access Threat Model=black-box2024.022
No AttackLLM Model=Guanaco-13B, Attack Version=Original2024.022
No AttackLLM Model=Vicuna-33B, Attack Version=Original2024.027
GCGLLM Model=Vicuna-33B, Attack Version=Original2024.0214
GCGLLM Model=Guanaco-13B, Attack Version=Original2024.0214
PAPLLM Model=Guanaco-13B, Attack Version=Original2024.0233
PAPLLM Model=Llama2-70B, Attack Version=Original2024.0244
PAPLLM Model=Vicuna-33B, Attack Version=Original2024.0252
GCG + PRPLLM Model=Vicuna-33B, Attack Version=PRP-W, Access Threat Model=white-box2024.0261
PAP + PRPLLM Model=Guanaco-13B, Attack Version=PRP-W, Access Threat Model=white-box2024.0270
GCG + PRPLLM Model=Vicuna-33B, Attack Version=PRP-B, Access Threat Model=black-box2024.0273
PAP + PRPLLM Model=Llama2-70B, Attack Version=PRP-B, Access Threat Model=black-box2024.0274
PAP + PRPLLM Model=Guanaco-13B, Attack Version=PRP-B, Access Threat Model=black-box2024.0274
PAP + PRPLLM Model=Llama2-70B, Attack Version=PRP-W, Access Threat Model=white-box2024.0276
GCG + PRPLLM Model=Guanaco-13B, Attack Version=PRP-W, Access Threat Model=white-box2024.0283
PAP + PRPLLM Model=Vicuna-33B, Attack Version=PRP-W, Access Threat Model=white-box2024.0284
PAP + PRPLLM Model=Vicuna-33B, Attack Version=PRP-B, Access Threat Model=black-box2024.0290
GCG + PRPLLM Model=Guanaco-13B, Attack Version=PRP-B, Access Threat Model=black-box2024.0291