The Feasibility of Topic-Based Watermarking on Academic Peer Reviews

TL;DR

This paper evaluates topic-based watermarking as a method to reliably attribute LLM-generated peer reviews without compromising review quality, demonstrating its robustness and practicality across various LLM configurations.

Contribution

It introduces and systematically assesses topic-based watermarking for peer review text, showing its effectiveness and robustness in real academic review scenarios.

Findings

TBW maintains review quality similar to non-watermarked text

TBW detection remains robust under paraphrasing

The method is effective across multiple LLM configurations

Abstract

Large language models (LLMs) are increasingly integrated into academic workflows, with many conferences and journals permitting their use for tasks such as language refinement and literature summarization. However, their use in peer review remains prohibited due to concerns around confidentiality breaches, hallucinated content, and inconsistent evaluations. As LLM-generated text becomes more indistinguishable from human writing, there is a growing need for reliable attribution mechanisms to preserve the integrity of the review process. In this work, we evaluate topic-based watermarking (TBW), a semantic-aware technique designed to embed detectable signals into LLM-generated text. We conduct a systematic assessment across multiple LLM configurations, including base, few-shot, and fine-tuned variants, using authentic peer review data from academic conferences. Our results show that TBW maintains review quality relative to non-watermarked outputs, while demonstrating robust detection performance under paraphrasing. These findings highlight the viability of TBW as a minimally intrusive and practical solution for LLM attribution in peer review settings.