Saved in:
Bibliographic Details
Main Authors: Li, Zhuang, Yi, Qiuping, Ji, Zongcheng, Lu, Yijian, Li, Yanqi, Xiao, Keyang, Liang, Hongliang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.12174
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915295918030848
author Li, Zhuang
Yi, Qiuping
Ji, Zongcheng
Lu, Yijian
Li, Yanqi
Xiao, Keyang
Liang, Hongliang
author_facet Li, Zhuang
Yi, Qiuping
Ji, Zongcheng
Lu, Yijian
Li, Yanqi
Xiao, Keyang
Liang, Hongliang
contents The rapid growth of Large Language Models (LLMs) raises concerns about distinguishing AI-generated text from human content. Existing watermarking techniques, like \kgw, struggle with low watermark strength and stringent false-positive requirements. Our analysis reveals that current methods rely on coarse estimates of non-watermarked text, limiting watermark detectability. To address this, we propose Bipolar Watermark (\tool), which splits generated text into positive and negative poles, enhancing detection without requiring additional computational resources or knowledge of the prompt. Theoretical analysis and experimental results demonstrate \tool's effectiveness and compatibility with existing optimization techniques, providing a new optimization dimension for watermarking in LLM-generated content.
format Preprint
id arxiv_https___arxiv_org_abs_2501_12174
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BiMarker: Enhancing Text Watermark Detection for Large Language Models with Bipolar Watermarks
Li, Zhuang
Yi, Qiuping
Ji, Zongcheng
Lu, Yijian
Li, Yanqi
Xiao, Keyang
Liang, Hongliang
Machine Learning
The rapid growth of Large Language Models (LLMs) raises concerns about distinguishing AI-generated text from human content. Existing watermarking techniques, like \kgw, struggle with low watermark strength and stringent false-positive requirements. Our analysis reveals that current methods rely on coarse estimates of non-watermarked text, limiting watermark detectability. To address this, we propose Bipolar Watermark (\tool), which splits generated text into positive and negative poles, enhancing detection without requiring additional computational resources or knowledge of the prompt. Theoretical analysis and experimental results demonstrate \tool's effectiveness and compatibility with existing optimization techniques, providing a new optimization dimension for watermarking in LLM-generated content.
title BiMarker: Enhancing Text Watermark Detection for Large Language Models with Bipolar Watermarks
topic Machine Learning
url https://arxiv.org/abs/2501.12174