LLM Watermarking Challenges with Reasoning Models

moyix · x · 2026-08-18

The author raises a technical question regarding the interaction between LLM watermarking and reasoning models. While the reasoning process (Chain of Thought) offers ample entropy for embedding watermarks, the user-visible final output contains significantly less entropy. This prompts consideration of whether watermarking can be effectively embedded or detected in the shorter, summarized outputs typical of reasoning models.

Related event: Watermarking LLMs Faces Challenges with Reasoning Models(2 posts)→

Original post →

More from Research

Research channel →