首个专为扩散语言模型设计的水印框架，解决非顺序生成带来的检测难题。
DMark: Order-Agnostic Watermarking for Diffusion Large Language Models
- 采用预测与双向依赖策略，恢复扩散模型水印可检测性
- 在1%误报率下检测率达92.0%-99.5%，远超旧方法的49.6%-71.2%
- 适用于对抗文本篡改，适合需可信生成内容的场景
扩散大语言模型（dLLMs）在保持与自回归模型相当质量的同时，生成速度更快，但现有水印方法因不适应其非顺序解码机制而失效。与自左至右生成标记的自回归模型不同，dLLMs可任意顺序确定标记，破坏了传统水印依赖的因果设计。本文提出DMark，首个专为dLLMs设计的水印框架。DMark引入三种互补策略：预测水印在实际上下文缺失时使用模型预测标记；双向水印利用扩散解码特有的正向与反向依赖；预测-双向水印结合两者以最大化检测强度。多组dLLMs实验表明，DMark在1%误报率下实现92.0%-99.5%的检测率，显著优于现有方法的49.6%-71.2%。同时对文本修改具备鲁棒性，证明非自回归语言模型水印可行。
原文摘要 · Abstract (English)
Diffusion large language models (dLLMs) offer faster generation than autoregressive models while maintaining comparable quality, but existing watermarking methods fail on them due to their non-sequential decoding. Unlike autoregressive models that generate tokens left-to-right, dLLMs can finalize tokens in arbitrary order, breaking the causal design underlying traditional watermarks. We present DMark, the first watermarking framework designed specifically for dLLMs. DMark introduces three complementary strategies to restore watermark detectability: predictive watermarking uses model-predicted tokens when actual context is unavailable; bidirectional watermarking exploits both forward and backward dependencies unique to diffusion decoding; and predictive-bidirectional watermarking combines both approaches to maximize detection strength. Experiments across multiple dLLMs show that DMark achieves 92.0-99.5% detection rates at 1% false positive rate while maintaining text quality, compared to only 49.6-71.2% for naive adaptations of existing methods. DMark also demonstrates robustness against text manipulations, establishing that effective watermarking is feasible for non-autoregressive language models.
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。