List your company

Harmful Input Detection on HarmBench

100MCA AccuracyReGA
40.7256.1171.586.89Jun 2, 2025

Evaluation results

MethodLinks
ReGAModel=qwen2025.0610091
ReGAModel=llama2025.068989
ReGAModel=vicuna2025.068884
ReGAModel=mistral2025.068663
ReGAModel=Average2025.068174
ReGAModel=baichuan2025.068080
ReGAModel=koala2025.064339