Fetching the paper…

Uncovering Safety Risks of Large Language Models through Concept Activation Vector · Around