A Nested Watermark for Large Language Models

The rapid advancement of large language models LLMs has raised concerns regarding their potential misuse, particularly in generating fake news and misinformation. To address these risks, watermarking techniques for autoregressive language models have emerged as a promising means for detecting...