Undetectable Watermarks for Language Models

arxiv.org

1 pointcryptohell1 comment

How many people do you think actually bother rephrasing LLM outputs? Do you personally ever copy paste a full chunk of a response?