Original title: More Criticism Does Not Make a Better Review: EquiReview-R
Zexing Zhang, Jichao Li, Tianyang Lei, Yude Fu +1 · September 2026
Artificial intelligence tools can now write long lists of critiques for research papers. However, having more criticism does not make a review better. An automated review can fail in two opposite ways: it can miss a serious flaw, or it can accuse the paper of a problem that the evidence does not actually support. To fix this, researchers created EquiReview-R. This system treats these two problems as separate risks. Instead of just writing text, it checks each criticism against evidence found in the paper to remove unsupported complaints. It then searches for missing issues from both independent and review-conditioned angles, before deciding whether to stop, continue, or defer. The researchers found that EquiReview-R cut major overcritique nearly in half, reduced missed major flaws, and stopped safely on over half the papers. The improvement came directly from revising existing critiques rather than just generating shorter output.
Imagine you're grading a student's science project. You might focus too much on minor neatness issues while missing that their hypothesis is fundamentally flawed. In this paper, the AI reviewer does something similar—it first checks if each criticism is supported by the project's data before looking for key errors it overlooked.
Peer reviewers rely on AI to handle the flood of papers, but flawed AI reviews can let bad research slip through or unfairly reject good work. This system makes AI reviews more accurate by fixing both missed problems and false alarms at once.
The abstract does not specify if the system generalizes to all academic fields or paper types. It relies on a constructed corpus for analysis, and real-world deployment challenges like computational cost or integration with existing workflows are not addressed.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
More Criticism Does Not Make a Better Review: EquiReview-R — Zexing Zhang, Jichao Li, Tianyang Lei et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.