pain-representation

This study explores whether large language models have distinct internal representations of pain, finds that they do, and examines the functional consequences through experiments, concluding with implications for AI safety and welfare.