Gamark: adaptive entropy-threshold watermarking via gaussian mixture modeling for large language models

Abstract

Existing generative watermarking methods rely on fixed or heuristic entropy thresholds in cross-task generation scenarios, leading to redundant watermark injection, detection noise, and degraded generation quality. To address these limitations, we propose GAMark, a Gaussian Mixture Model (GMM)-based semantics-aware Adaptive entropy threshold waterMarking framework. Motivated by the manifold hypothesis in the logits space, GAMark adopts an offline semantic modeling and online frozen inference par

Categories

@article{zhao2026gamark,
  title = {{Gamark: adaptive entropy-threshold watermarking via gaussian mixture modeling for large language models}},
  author = {Jing Zhao and Hong-Wei Yang and Heng-Ji Dong and Hui He and Wei-Zhe Zhang},
  year = {2026},
  month = sep,
  journal = {Cybersecurity},
  doi = {10.1186/s42400-026-00607-1},
  url = {https://www.semanticscholar.org/paper/745bb40ffc9b6659100956d864c90751dc31e01a}
}