ResearchBenchmarksJailbreaking Attack Detection on Role-Playing Jailbreaking PromptsFollow71AccMCAReGA3.420.9538.556.05Jun 2, 2025Evaluation ResultsMethodMethodLinksAccMCAAccMFPReGATarget LLM=llamaTarget LLM=llama2025.067171ReGATarget LLM=vicunaTarget LLM=vicuna2025.066944ReGATarget LLM=qwenTarget LLM=qwen2025.066241ReGATarget LLM=mistralTarget LLM=mistral2025.065136ReGATarget LLM=baichuanTarget LLM=baichuan2025.063030ReGATarget LLM=koalaTarget LLM=koala2025.0666