Quillbot

Watermarking text generated by large language models (LLMs) is more difficult than it appears, according to a technical analysis. Current methods, which force models to select tokens from a 'green-lis…