ResearchBenchmarksSafety Detection on Alpaca and AdvBench Conversation Level (test)Safety Detection on Alpaca and AdvBench Conversation Level (test)Follow100MCA AccuracyReGA90.6493.0795.597.93Jun 2, 2025Evaluation resultsMethodMethodLinksMCA AccuracyMFP AccuracyAUROCReGAModel=llamaModel=llama2025.06100100100ReGAModel=mistralModel=mistral2025.06100100100ReGAModel=qwenModel=qwen2025.0699100100ReGAModel=baichuanModel=baichuan2025.069899100ReGA (Average)Model=AverageModel=Average2025.06979298.5ReGAModel=vicunaModel=vicuna2025.06959498.4ReGAModel=koalaModel=koala2025.06916192.5