XMark: Reliable Multi-Bit Watermarking for LLM-Generated Texts

arXiv cs.CL / 4/8/2026

💬 OpinionIdeas & Deep AnalysisTools & Practical UsageModels & Research

共有:

Key Points

XMark is a proposed multi-bit watermarking method for LLM-generated text that aims to embed imperceptible binary messages for reliable attribution and tracing of malicious LLM usage.
The work targets shortcomings of prior approaches, including computational infeasibility for large messages, weak quality–decoding trade-offs, and sharp decoding accuracy degradation when only a limited number of tokens are available.
XMark’s encoder is designed to produce a less distorted logit distribution to preserve text quality, while a tailored decoder is built to recover the message reliably under token-limited conditions.
Experiments across multiple downstream tasks reportedly show XMark improves decoding accuracy substantially while maintaining watermark text quality and outperforming prior methods.
The authors provide an implementation at the linked GitHub repository, supporting replication and further evaluation.

Abstract

Multi-bit watermarking has emerged as a promising solution for embedding imperceptible binary messages into Large Language Model (LLM)-generated text, enabling reliable attribution and tracing of malicious usage of LLMs. Despite recent progress, existing methods still face key limitations: some become computationally infeasible for large messages, while others suffer from a poor trade-off between text quality and decoding accuracy. Moreover, the decoding accuracy of existing methods drops significantly when the number of tokens in the generated text is limited, a condition that frequently arises in practical usage. To address these challenges, we propose \textsc{XMark}, a novel method for encoding and decoding binary messages in LLM-generated texts. The unique design of \textsc{XMark}'s encoder produces a less distorted logit distribution for watermarked token generation, preserving text quality, and also enables its tailored decoder to reliably recover the encoded message with limited tokens. Extensive experiments across diverse downstream tasks show that \textsc{XMark} significantly improves decoding accuracy while preserving the quality of watermarked text, outperforming prior methods. The code is at https://github.com/JiiahaoXU/XMark.