Pairwise watermarking improves detection in machine generated text
TTMark: Pairwise Distortion-Free Watermarking Beyond Single-Token Entropy
Cryptography and SecurityComputation and Language
Summary
Watermarks in AI-written text help show who created it without changing the text itself. Existing watermarks check one word at a time, limiting how well they can be found. The authors introduce a method that watermarks pairs of words together, making detection much stronger without altering the text quality. This new approach also works better if the text is edited and can find watermarks in smaller excerpts more reliably.
What this means in practice
- •For content moderation teams: Identify AI-generated text more reliably, even after edits, to better police generated content on platforms.
- •For digital rights managers: Embed stronger watermarks in AI-generated writing to prove ownership without degrading text quality.
Authors
Ruibo Chen, Zhengmian Hu, Donghang Lu, Xuehao Cui, Georgios Milis, Yihan Wu, Jian Du, Heng Huang
Abstract
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.