Selective Disclosure Watermarking for Large Language Models

2026-07-06

AI Summary

This work addresses the privacy risks in existing multi-bit watermarking methods for large language models, which lack selective disclosure capabilities and require revealing all embedded information to verify any portion of the watermark. To overcome this limitation, we propose Hierarchical Vocabulary Routing (HeRo), a novel framework that recursively partitions the vocabulary and distributes watermarks across a multi-level structure. HeRo enables fine-grained access control by allowing verifiers at different authorization levels to decode only the metadata corresponding to their permissions, while preserving the fluency and quality of generated text. Experimental results demonstrate that HeRo achieves high detection accuracy and low inference latency, effectively supporting permission-based hierarchical watermark decoding without compromising output integrity.

Abstract

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.

Problem

• selective disclosure
• watermarking
• large language models
• privacy
• metadata

Innovation

• Selective Disclosure
• Hierarchical Vocabulary Routing
• Watermarking
• Large Language Models
• Access Control