A0246
Title: Segmenting watermarked texts from language models
Authors: Xianyang Zhang - Texas A\&M University (United States) [presenting]
Guanxun Li - Texas A&M University (United States)
Xingchi Li - Texas A&M University (United States)
Abstract: Watermarking is a technique that involves embedding (nearly) unnoticeable statistical signals within generated content to help trace its source. The focus is on a scenario where an untrusted third-party user sends prompts to a trusted language model (LLM) provider, who then generates a text from their LLM with a watermark. This setup makes it possible for a detector to later identify the source of the text if the user publishes it. The user can modify the generated text by substitutions, insertions, or deletions. The objective is to develop a statistical method to detect if a published text is LLM-generated, mainly from the perspective of a detector. A methodology is further proposed to segment the published text into watermarked and non-watermarked sub-strings. The proposed approach is built upon randomization tests and change point detection techniques. The demonstrated method ensures Type I and Type II error control and can accurately identify watermarked sub-strings by finding the corresponding change point locations. To validate the technique, it is applied to texts generated by several language models with prompts extracted from Google's C4 dataset and obtain encouraging numerical results.
