List your company

Safety Representation Extraction on Prompt level dataset

100MCA AccuracyReGA
95.8496.929899.08Jun 2, 2025

Evaluation results

MethodLinks
ReGAModel=llama-13B2025.0610010099.2
ReGAModel=qwen-14B2025.061009899.5
ReGAModel=vicuna-13B2025.06969498