ResearchBenchmarksSafety Detection on Alpaca and AdvBench Prompt Level (test)Follow100Accuracy (MCA)ReGA91.6893.849698.16Jun 2, 2025Evaluation ResultsMethodMethodLinksAccuracy (MCA)Accuracy (MFP)AUROCReGAModel=llamaModel=llama2025.0610010099.6ReGAModel=qwenModel=qwen2025.06999999.8ReGAModel=mistralModel=mistral2025.069910099.8ReGA (Average)Model=AverageModel=Average2025.06979197.5ReGAModel=baichuanModel=baichuan2025.06969098.5ReGAModel=vicunaModel=vicuna2025.06959498.3ReGAModel=koalaModel=koala2025.06926189