Saved in:
Bibliographic Details
Main Authors: Ai, Zhenxin, He, Haiyun
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.10977
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910249474064384
author Ai, Zhenxin
He, Haiyun
author_facet Ai, Zhenxin
He, Haiyun
contents Watermarking for large language models (LLMs) is a promising approach for detecting LLM-generated text and enabling responsible deployment. However, existing watermarking methods are often vulnerable to semantic-invariant attacks, such as paraphrasing. We propose PASA, a principled, robust, and distortion-free watermarking algorithm that embeds and detects a watermark at the semantic level. PASA operates on semantic clusters in a latent embedding space and constructs a distributional dependency between token and auxiliary sequences via shared randomness synchronized by a secret key and semantic history. This design is grounded in our theoretical framework that characterizes a jointly optimal embedding-detection pair, achieving the fundamental trade-offs among detection accuracy, robustness, and distortion. Evaluations across multiple LLMs and semantic-invariant attacks demonstrate that PASA remains robust even under strong paraphrasing attacks while preserving high text quality, outperforming standard vocabulary-space baselines. Ablation studies further validate the effectiveness of our hyperparameter choices. Webpage: https://ai-kunkun.github.io/PASA_page/.
format Preprint
id arxiv_https___arxiv_org_abs_2605_10977
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PASA: A Principled Embedding-Space Watermarking Approach for LLM-Generated Text under Semantic-Invariant Attacks
Ai, Zhenxin
He, Haiyun
Cryptography and Security
Artificial Intelligence
Watermarking for large language models (LLMs) is a promising approach for detecting LLM-generated text and enabling responsible deployment. However, existing watermarking methods are often vulnerable to semantic-invariant attacks, such as paraphrasing. We propose PASA, a principled, robust, and distortion-free watermarking algorithm that embeds and detects a watermark at the semantic level. PASA operates on semantic clusters in a latent embedding space and constructs a distributional dependency between token and auxiliary sequences via shared randomness synchronized by a secret key and semantic history. This design is grounded in our theoretical framework that characterizes a jointly optimal embedding-detection pair, achieving the fundamental trade-offs among detection accuracy, robustness, and distortion. Evaluations across multiple LLMs and semantic-invariant attacks demonstrate that PASA remains robust even under strong paraphrasing attacks while preserving high text quality, outperforming standard vocabulary-space baselines. Ablation studies further validate the effectiveness of our hyperparameter choices. Webpage: https://ai-kunkun.github.io/PASA_page/.
title PASA: A Principled Embedding-Space Watermarking Approach for LLM-Generated Text under Semantic-Invariant Attacks
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2605.10977