Fast segmentation of watermarked texts from large language models through an epidemic change-point framework
stat.ML, cs.LG, stat.AP, stat.ME
Submitted: 2025-09-25
Updated: 2026-07-08
Comments: 68 pages
License: http://creativecommons.org/licenses/by/4.0/
The gist: With the growing use of large language models, concerns over content authenticity have spurred a variety of watermarking schemes.
Terminology
Abstract
With the growing use of large language models, concerns over content authenticity have spurred a variety of watermarking schemes. These schemes use secret keys to detect machine-generated text while remaining imperceptible to readers. Detection typically reduces to statistical hypothesis testing for the presence of watermarks, a topic that is now well studied. In contrast, the finer-grained task of localizing which segments of a text are watermarked is much less explored; existing approaches often lack scalability or guarantees robust to paraphrasing and post-editing. We bring a new perspective to this segmentation problem through the lens of epidemic change-points and, by exploiting this connection, propose WISER, a novel and computationally efficient watermark segmentation algorithm. We establish finite-sample error bounds and consistency for detecting multiple watermarked segments in a single text. Complementing these theoretical results, our extensive numerical experiments show that WISER outperforms state-of-the-art baseline methods, both in terms of computational speed as well as accuracy, on various benchmark datasets embedded with diverse watermarking schemes. Together, these theoretical and empirical results position WISER as an effective tool for watermark localization and illustrate how classical statistical ideas can yield theoretically valid and computationally efficient solutions to a modern problem of immediate importance.
Sources
- GPT-4 Technical Report
- Detecting Fake News Using Machine Learning : A Systematic Literature Review
- BigScience: A Case Study in the Social Construction of a Multilingual Large Language Model
- A Statistical Hypothesis Testing Framework for Data Misappropriation Detection in Large Language Models
- Towards Better Statistical Understanding of Watermarking LLMs
- Can LLM-Generated Misinformation Be Detected?
- Measuring Association on Topological Spaces Using Kernels and Geometric Graphs
- GLTR: Statistical Detection and Visualization of Generated Text
- Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text
- StarCoder: may the source be with you!
- Adaptive Testing for Segmenting Watermarked Texts From Language Models
- Optimal Estimation of Watermark Proportions in Hybrid AI-Human Texts
- Monitoring AI-Modified Content at Scale: A Case Study on the Impact of ChatGPT on AI Conference Peer Reviews
- A more efficient algorithm to compute the Rand Index for change-point problems
- Evaluating Durability: Benchmark Insights into Multimodal Watermarking
- A Training-free Method for LLM Text Attribution
- Release Strategies and the Social Impacts of Language Models
- LLaMA: Open and Efficient Foundation Language Models
- HowkGPT: Investigating the Detection of ChatGPT-generated University Student Homework through Context-Aware Perplexity Analysis
- Large Language Models for Education: A Survey and Outlook
Related papers
- Uniform-in-Time Weak Propagation-of-Chaos in Shallow Neural Networks
- Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors
- Quantifying Error Propagation and Model Collapse in Diffusion Models
- Behavior of prediction performance metrics with rare events
- Statistical Inference and Learning for Shapley Additive Explanations (SHAP)
- Control Variate Design for Score-Function Gradient Estimators in Importance-Weighted Variational Inference