Verified, then logged.
Sun 22 Nov 2026 · 00:17 KST / 15:17 UTC한국어 | ENSearch /

Breaking01 / AI

OmniReview offers benchmark for reviewer recommendations

Researcher profiles and staged evaluation aim to reflect real editorial workflows.

An academic editor viewed from behind compares a manuscript with researcher expertise profiles across two monitors to select potential peer reviewers. AI-generated image

The OmniReview researchers proposed a benchmark for recommending academic peer reviewers in an arXiv revision published on 11 Nov 2026. They reported that Pro-MMoE, the framework presented alongside it, achieved the best performance on most evaluation metrics.

According to the abstract, the dataset combines information from several academic platforms. The researchers described a process for distinguishing researcher identities and collecting verified review records. Their evaluation framework spans candidate retrieval through to the identification of suitable experts.

The researchers said Pro-MMoE combines large language models with multi-task learning, which trains a model on several tasks together. It uses language-model-generated researcher profiles to preserve detailed expertise information and make recommendations easier to interpret. Its task-adaptive architecture is designed to balance conflicting evaluation goals.

The researchers identified a shortage of large, verified benchmarks as a limitation of earlier work. They argued that simplified metrics fail to reflect editorial workflows. They also described information bottlenecks and limited interpretability in existing embedding methods, which compress meaning into numerical representations.

The abstract does not disclose performance figures for individual metrics.