List your company

Jailbreak Detection on DrAttack

99AccuracyGradSafe
-3.9622.7749.576.23Dec 3, 2025

Evaluation results

MethodLinks
GradSafeTarget LLM=Llama2-7B2025.129999
IMAGTarget LLM=Vicuna-13B2025.129899
PPLTarget LLM=Llama3-8B2025.129798
IMAGTarget LLM=Llama3-8B2025.129698
IMAGTarget LLM=Mistral-7B2025.128692
LlamaGTarget LLM=Vicuna-13B2025.128592
LlamaGTarget LLM=Mistral-7B2025.128491
LlamaGTarget LLM=Vicuna-7B2025.128491
IMAGTarget LLM=Vicuna-7B2025.128491
IMAGTarget LLM=Llama2-7B2025.126579
LlamaGTarget LLM=Llama2-7B2025.125772
LlamaGTarget LLM=Llama3-8B2025.125571
Self-ExTarget LLM=Mistral-7B2025.125167
Self-ExTarget LLM=Llama2-7B2025.122427
Self-ExTarget LLM=Llama3-8B2025.121830
OAPITarget LLM=Mistral-7B2025.1247
OAPITarget LLM=Vicuna-7B2025.1247
OAPITarget LLM=Vicuna-13B2025.1248
Self-ExTarget LLM=Vicuna-7B2025.1236
OAPITarget LLM=Llama2-7B2025.1237
OAPITarget LLM=Llama3-8B2025.1236
PPLTarget LLM=Mistral-7B2025.1200
GradSafeTarget LLM=Mistral-7B2025.1200
PPLTarget LLM=Vicuna-7B2025.1200
GradSafeTarget LLM=Vicuna-7B2025.1200
PPLTarget LLM=Vicuna-13B2025.1200
Self-ExTarget LLM=Vicuna-13B2025.1200
GradSafeTarget LLM=Vicuna-13B2025.1200
PPLTarget LLM=Llama2-7B2025.1200
GradSafeTarget LLM=Llama3-8B2025.1200