TTMark：超越单令牌熵的成对无失真水印

中文摘要 AI 辅助

无失真水印能够在保持输出分布的同时，实现对机器生成文本的可靠归属。然而，现有方法对每个生成的令牌独立操作，其检测能力从根本上受限于下一令牌分布的熵。我们提出了串联令牌水印（TTMARK），一个通用的成对水印框架，将无失真水印从单个令牌扩展到相邻令牌对。通过对连续令牌的联合分布进行水印处理，TTMARK将有效水印字母表从V扩大到$V^2$，使检测器能够同时利用令牌熵和条件熵，同时在联合分布上保持无失真性。我们进一步引入了一种分支隔离的串联生成算法，该算法在单次前向传播中高效地构建联合分布。理论上，我们证明了成对水印在低熵区域能够实现更好的期望检测强度。跨多个语言模型、数据集和三种代表性无失真水印方案的广泛实验表明，TTMARK在不降低生成质量的情况下持续提高可检测性，同时增强了对编辑的鲁棒性，并显著提升了局部水印检测能力。

英文摘要

Distortion-free watermarking enables reliable attribution of machine-generated text while preserving output distribution. However, existing methods operate independently on each generated token, making their detection capability fundamentally constrained by the entropy of the next-token distribution. We present Tandem Token WaterMark (TTMARK), a general pairwise watermarking framework that extends distortion-free watermarking from individual tokens to adjacent token pairs. By watermarking the joint distribution of consecutive tokens, TTMARK enlarges the effective watermarking alphabet from V to $V^2$, allowing the detector to exploit both token entropy and conditional entropy while preserving distortion-freeness over the joint distribution. We further introduce a branch-isolating concatenated tandem generation algorithm that efficiently constructs the joint distribution in a single forward pass. Theoretically, we show that pairwise watermarking achieves better expected detection strength in low-entropy regimes. Extensive experiments across multiple language models, datasets, and three representative distortion-free watermarking schemes demonstrate that TTMARK consistently improves detectability without degrading generation quality, while also improving robustness to edits and substantially enhancing localized watermark detection.

发表机构

• University of Maryland, College Park（马里兰大学帕克分校）
• TikTok

机构由 AI 辅助整理，请以论文原文为准。