paperAugust 2026Unreviewed

SemTrace: Source-Grounded Semantic Signatures for Tracing LLM Exposure to Protected Documents

Jun-Yan Zhang, Yuan Zeng, Yong-Wei Huang, Zu-Hao Ouyang, Hong Chen, Xuming Hu

Abstract

Large language models are increasingly used to read documents and produce downstream text, creating a provenance problem when the document owner cannot control or inspect the model that performs the generation. We introduce SemTrace, a source-grounded semantic watermark for detecting whether a generated review was influenced by a known protected manuscript copy. Rather than biasing token probabilities or imposing surface-form patterns, SemTrace constructs a document-specific binary signature fro

Cite

@misc{zhang2026semtrace,
  title = {{SemTrace: Source-Grounded Semantic Signatures for Tracing LLM Exposure to Protected Documents}},
  author = {Jun-Yan Zhang and Yuan Zeng and Yong-Wei Huang and Zu-Hao Ouyang and Hong Chen and Xuming Hu},
  year = {2026},
  month = aug,
  eprint = {2608.29575},
  archivePrefix = {arXiv},
  url = {https://www.semanticscholar.org/paper/d864a78fe762e931a96efe5ea38bdf249372cacd}
}