Our work “PaCE: Parsimonious Concept Engineering for Large Language Models” has been accpeted to NeurIPS 2024!