Stability-Aware Feature Design for Robust Watermark Detection in Machine-Generated Text

arXiv:2608.181022026-08-20

The widespread adoption of large language models (LLMs) has intensified the demand for principled methods to distinguish human from machine-generated text. Watermarking provides a promising avenue, yet existing detectors exhibit sharp performance deterioration under multiple paraphrasing and when applied to shorter texts. We introduce Pattern Stability Score (PSS), a novel detection framework that leverages local statistical features and stability

Authors · Sina Mansouri, Mohit Marvania, Abolfazl Safikhani