RepreGuard: Detecting LLM-Generated Text by Revealing Hidden Representation Patterns

Published in Transactions of the Association for Computational Linguistics (TACL), 2025

Abstract: