Evaluate Bias without Manual Test Sets: A Concept Representation Perspective for LLMs

Open in new window