From TDMA to CDMA: A Multi-bit Watermark for Diffusion Language Models
Baizhou Huang, Xiaojun Wan
摘要
While DLMs have emerged as an alternative to ARMs, robust content provenance mechanisms for this architecture remain unexplored. Existing multi-bit watermarking schemes, heavily reliant on the sequential generation of ARMs, cannot be directly applied to DLMs. In this paper, we reframe the multi-bit watermarking problem through a novel Digital Signal Processing (DSP) lens. We draw an analogy between prior works and TDMA (Time Division Multiple Access), revealing their inherent limitations. To overcome these limitations, we introduce CDMArk, the first multi-bit watermarking framework tailored for DLMs, orchestrating a paradigm shift from TDMA to CDMA (Code Division Multiple Access). CDMArk encodes the entire watermark message across all token positions holographically. We further provide rigorous statistical guarantees for its detection process. Extensive experiments demonstrate that CDMArk achieves a new state-ofthe-art Pareto frontier between imperceptibility and effectiveness.
问问这篇 Paper
智能体会读完全文。
Lune 把这篇 Paper 索引到了每一个公式,引用它的顶会 Paper 也一样。你提问,回答直接引用原文。
它引用的顶会 Paper15
- Structured Denoising Diffusion Models in Discrete State-SpacesJacob Austin, Daniel D. Johnson, Jonathan Ho, Daniel Tarlow 等NeurIPS 2021 · 被引用 2,256 次
- Large Language Diffusion ModelsShen Nie, Fengqi Zhu, Zebin You, Xiaolu Zhang 等NeurIPS 2025 · 被引用 949 次
- Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defenseKalpesh Krishna, Yixiao Song, Marzena Karpinska, John Wieting 等NeurIPS 2023 · 被引用 657 次
- Discrete Diffusion Modeling by Estimating the Ratios of the Data DistributionAaron Lou, Chenlin Meng, Stefano ErmonICML 2024 · 被引用 473 次
- LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion ModelsFengqi Zhu, Rongzhen Wang, Shen Nie, Xiaolu Zhang 等ACL 2026 · 被引用 229 次
相关 Paper
- Watermarking Diffusion Language ModelsThibaud Gloaguen, Robin Staab, Nikola Jovanović, Martin VechevICLR 2026 · 被引用 13 次
- dgMARK: Decoding-Guided Watermarking for Diffusion Language ModelsPyo Min Hong, Albert NoICML 2026
- You Can Have a Second Chance: Unbiased and Multi-bit Watermarking for Diffusion Language Models with Regret-based RemaskingKe Yang, Dongyang Liang, Jing Yu, Shuguang Yuan 等ACL 2026
- XMark: Reliable Multi-Bit Watermarking for LLM-Generated TextsJiahao Xu, Rui Hu, Olivera Kotevska, Zikai ZhangACL 2026 · 被引用 1 次
- IPMark: A Sentence-Level Watermark for LLMs with Hierarchical Personalization and Efficient DetectionWenbo An, Lianwei Wu, Zehao WangICML 2026
