提出可自适应分布的LLM水印框架，兼顾检测率与文本质量。
Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approach
- 统一优化水印生成与检测，理论推导出最优解
- 在极低误报率下仍保持高检测率，且无文本失真
- 适合需高可靠性水印的AI内容审核场景
水印已成为区分AI生成文本与人工文本的关键方法。现有方法通常缺乏形式化最优性保证，或分别设计水印方案与检测器。本文提出一种统一的理论框架，联合优化大型语言模型（LLM）的水印方案与检测器。目标是在最大化检测性能的同时，控制最坏情况下的误报率（FPR）和文本质量失真。我们推导出该联合设计的闭式最优解，并刻画了水印可检测性与失真之间的根本权衡。特别地，揭示最优水印方案应适配LLM的生成分布。基于理论洞察，我们提出一种无失真的分布自适应水印算法（DAWA），利用代理模型实现模型无关性和高效性。在Llama2-13B和Mistral-8×7B模型上的实验验证了该方法的有效性，尤其在超低误报率下表现优异。代码已开源：https://github.com/yepengliu/DAWA。
原文摘要 · Abstract (English)
Watermarking has emerged as a crucial method to distinguish AI-generated text from human-created text. Current watermarking approaches often lack formal optimality guarantees or address the scheme and detector design separately. In this paper, we introduce a novel, unified theoretical framework for watermarking Large Language Models (LLMs) that jointly optimizes both the watermarking scheme and detector. Our approach aims to maximize detection performance while maintaining control over the worst-case false positive rate (FPR) and distortion on text quality. We derive closed-form optimal solutions for this joint design and characterize the fundamental trade-off between watermark detectability and distortion. Notably, we reveal that the optimal watermarking schemes should be adaptive to the LLM's generative distribution. Building on our theoretical insights, we propose a distortion-free, distribution-adaptive watermarking algorithm (DAWA) that leverages a surrogate model for model-agnosticism and efficiency. Experiments on Llama2-13B and Mistral-8$\times$7B models confirm the effectiveness of our approach, particularly at ultra-low FPRs. Our code is available at https://github.com/yepengliu/DAWA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。