Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoption

Yepeng Liu, Xuandong Zhao, Dawn Song +2

Despite progress in watermarking algorithms for large language models (LLMs), real-world deployment remains limited. We argue that this gap stems from misaligned incentives among L…