Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio

Georgios Milis, Yubin Qin, Yihan Wu, Heng Huang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:88821-88836, 2026.

Abstract

As policy catches up with the capabilities of generative AI, watermarking is central to content provenance efforts. Inference-time watermarks for autoregressive models are unfit for continuous modalities due to discretization inconsistencies. Existing methods overcome this by finetuning the modality tokenizers, nullifying the watermark’s training-free advantage. In this work, motivated by the vocabulary redundancy of discretization, we propose an elegant solution for powerful and robust watermarking of synthetic audio. We theoretically analyze the impact of token errors on watermark detection, and effectively mitigate them using a reduced vocabulary obtained via community detection. Thorough experiments showcase that our gradient-free method can boost detectability by several orders of magnitude, while also achieving built-in robustness to audio modifications. Broadly, we discover a new state-of-the-art for token-level watermarks in multimedia, which simply arises from the nature of discrete representation learning.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-milis26a, title = {Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio}, author = {Milis, Georgios and Qin, Yubin and Wu, Yihan and Huang, Heng}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {88821--88836}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/milis26a/milis26a.pdf}, url = {https://proceedings.mlr.press/v306/milis26a.html}, abstract = {As policy catches up with the capabilities of generative AI, watermarking is central to content provenance efforts. Inference-time watermarks for autoregressive models are unfit for continuous modalities due to discretization inconsistencies. Existing methods overcome this by finetuning the modality tokenizers, nullifying the watermark’s training-free advantage. In this work, motivated by the vocabulary redundancy of discretization, we propose an elegant solution for powerful and robust watermarking of synthetic audio. We theoretically analyze the impact of token errors on watermark detection, and effectively mitigate them using a reduced vocabulary obtained via community detection. Thorough experiments showcase that our gradient-free method can boost detectability by several orders of magnitude, while also achieving built-in robustness to audio modifications. Broadly, we discover a new state-of-the-art for token-level watermarks in multimedia, which simply arises from the nature of discrete representation learning.} }
Endnote
%0 Conference Paper %T Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio %A Georgios Milis %A Yubin Qin %A Yihan Wu %A Heng Huang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-milis26a %I PMLR %P 88821--88836 %U https://proceedings.mlr.press/v306/milis26a.html %V 306 %X As policy catches up with the capabilities of generative AI, watermarking is central to content provenance efforts. Inference-time watermarks for autoregressive models are unfit for continuous modalities due to discretization inconsistencies. Existing methods overcome this by finetuning the modality tokenizers, nullifying the watermark’s training-free advantage. In this work, motivated by the vocabulary redundancy of discretization, we propose an elegant solution for powerful and robust watermarking of synthetic audio. We theoretically analyze the impact of token errors on watermark detection, and effectively mitigate them using a reduced vocabulary obtained via community detection. Thorough experiments showcase that our gradient-free method can boost detectability by several orders of magnitude, while also achieving built-in robustness to audio modifications. Broadly, we discover a new state-of-the-art for token-level watermarks in multimedia, which simply arises from the nature of discrete representation learning.
APA
Milis, G., Qin, Y., Wu, Y. & Huang, H.. (2026). Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:88821-88836 Available from https://proceedings.mlr.press/v306/milis26a.html.

Related Material