Selective Disclosure Watermarking for Large Language Models

Source

Source name: arXiv
Source kind: paper
Authors / organization: Xuyang Chen, Xiang Li, Yangxinyu Xie, Qi Long
Venue: arXiv: cs.CR
Published: 2026-07-06
Collected: 2026-07-07
URL: http://arxiv.org/abs/2607.05353v1
PDF: https://arxiv.org/pdf/2607.05353v1
DOI: UNKNOWN

Plain-Language Brief

Researchers have proposed a new watermarking method for large language models (LLMs) that allows selective disclosure of embedded metadata. This means that the method can control which parts of the embedded message are revealed, reducing unnecessary information exposure and addressing privacy concerns.

What the Document Says

The text discusses a watermarking framework called Hierarchical Vocabulary Routing (HeRo) for large language models.
HeRo allows selective disclosure of embedded metadata, meaning different verifiers can decode only the portions of the payload corresponding to their access level.
The method preserves the unbiasedness of the underlying sampling process and maintains text quality.
Experiments show that HeRo supports fine-grained access control while achieving high detection accuracy and low latency.
The code for this framework is available on GitHub.

Key Concepts Mentioned

watermarking
large language models (LLMs)
selective disclosure
hierarchical vocabulary routing (HeRo)
access control
detection accuracy
latency

Claims Made by the Document

HeRo allows selective disclosure of embedded metadata.
The method preserves the unbiasedness of the underlying sampling process and maintains text quality.
Experiments demonstrate that HeRo supports fine-grained access control while achieving high detection accuracy and low latency.

Evidence Used

The text describes how HeRo works, including its ability to selectively disclose metadata and preserve text quality.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Watermarking methods embed imperceptible and verifiable signals into text generated by large language models (LLMs). Existing approaches include zero-bit schemes for distinguishing synthetic text from human writing and multi-bit schemes for embedding metadata. However, current multi-bit watermarking methods do not allow selective disclosure: verifying any part of the watermark requires revealing the entire embedded message. This lack of control leads to unnecessary information exposure and raises privacy concerns. We propose Hierarchical Vocabulary Routing (HeRo), a watermarking framework that enables selective disclosure of embedded metadata. The method recursively partitions the vocabulary and distributes watermark information across hierarchical layers, so that different verifiers can decode only the portions of the payload corresponding to their access level. We show that the proposed scheme preserves the unbiasedness of the underlying sampling process and thus maintains text quality. Experiments demonstrate that our framework supports fine-grained access control while achieving high detection accuracy and low latency. Code is available at https://github.com/xuyangc03/hero-watermark.