Robust and Efficient Watermarking Framework for Safeguarding Intellectual Property in Large Language Models

REMARK-LLM, a novel watermarking framework, efficiently embeds binary signatures into texts generated by large language models (LLMs) while preserving semantic integrity and exhibiting resilience against watermark detection and removal attacks.