ByteMark: A Unified Byte-Level Watermark for Detection Across Language Models
Abstract
As watermarking of LLM-generated text enters deployment, users increasingly need to check text whose source is unknown. Existing detectors read the watermark through a specific tokenizer or embedding pipeline, and even if all providers adopted the same watermarking method, differences in these representations would leave their watermarks incompatible, and the text alone does not reveal which representation produced it. We propose ByteMark, a unified watermarking scheme that defines its signal on the UTF-8 bytes of published text. Models with different tokenizers embed this signal under one public specification and key, and a single detector verifies the text from the text and key alone. When the tokenizer or embedding model of a baseline detector is replaced, its detection degrades sharply, whereas ByteMark, which uses neither, retains its detection rate; ByteMark also remains detectable when participating models continue or rewrite the text.
Related
- Predictive Likelihood Ratios for Language Model Watermark Detection
- Optimal Watermark Localization in Mixed-Source Large Language Model Texts
- On the Multi-Index, Multi-Rate, and Multi-Phase Dynamics of Decoder-Only Language Models: A Unified Hybrid Framework for Generative and Agentic Systems
- FissionGuard: Testing Local Watermark Proportions with FDR Control
- CountScan: Simultaneous Lower Bounds on Historical Watermark Matches over All Text Intervals
BibTeX
@misc{wang2026bytemark,
title = {{ByteMark: A Unified Byte-Level Watermark for Detection Across Language Models}},
author = {Xiaopu Wang and Runze Li},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.14427},
eprint = {2610.14427},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.14427},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - ByteMark: A Unified Byte-Level Watermark for Detection Across Language Models
AU - Xiaopu Wang
AU - Runze Li
PY - 2026
DA - 2026/10/10
PB - arXiv
DO - 10.48550/arxiv.2610.14427
UR - https://arxiv.org/abs/2610.14427
AB - As watermarking of LLM-generated text enters deployment, users increasingly need to check text whose source is unknown. Existing detectors read the watermark through a specific tokenizer or embedding pipeline, and even if all providers adopted the same watermarking method, differences in these representations would leave their watermarks incompatible, and the text alone does not reveal which representation produced it. We propose ByteMark, a unified watermarking scheme that defines its signal on the UTF-8 bytes of published text. Models with different tokenizers embed this signal under one public specification and key, and a single detector verifies the text from the text and key alone. When the tokenizer or embedding model of a baseline detector is replaced, its detection degrades sharply, whereas ByteMark, which uses neither, retains its detection rate; ByteMark also remains detectable when participating models continue or rewrite the text.
ER -