On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists

With the advancement of AI capabilities, AI reviewers are beginning to be deployed in scientific peer review, yet their capability and credibility remain in question: many scientists simply view them as probabilistic systems without the expertise to evaluate research, while other researchers are more optimistic about their readiness without concrete evidence. Understanding what AI reviewers do well, where they fall short, and what challenges remain is essential. However, existing evaluations of AI reviewers have focused on whether their verdicts match human verdicts (e.g., score alignment, acceptance prediction), which is insufficient to characterize their capabilities and limits. In this paper, we close this gap through a large-scale expert annotation study, in which 45 domain scientists in Physical, Biological, and Health Sciences spent 469 hours rating 2,960 individual criticisms (each targeting one specific aspect of a paper) from human-written and AI-generated reviews of 82 Nature-family papers on correctness, significance, and sufficiency of evidence. On a composite of all three dimensions, a reviewing agent powered by GPT-5.2 scores above each paper's top-rated human reviewer (60.0% vs. 48.2%, p = 0.009), while all three AI reviewers (including Gemini 3.0 Pro and Claude Opus 4.5) exceed the lowest-rated human across every dimension. AI reviewers' accurate criticisms are also more often rated significant and well-evidenced, and surface a distinct 26% of issues no human raises. However, AI reviewers overlap far more than humans do (21% vs. 3% for cross-reviewer pairs), and exhibit 16 recurring weaknesses humans do not share, such as limited subfield knowledge, lack of long context management over multiple files, and overly critical stance on minor issues. Overall, our results position current AI reviewers as complements to, not substitutes for, human reviewers.

Predicting Chemical Reaction Outcomes Based on Electron Movements Using Machine Learning

Accurately predicting chemical reaction outcomes and potential byproducts is a fundamental task of modern chemistry, enabling the efficient design of synthetic pathways and driving progress in chemical science. Reaction mechanism, which tracks electron movements during chemical reactions, is critical for understanding reaction kinetics and identifying unexpected products. Here, we present Reactron, the first electron-based machine learning model for general reaction prediction. Reactron integrates electron movement into its predictions, generating detailed arrow-pushing diagrams that elucidate each mechanistic step leading to product formation. We demonstrate the high predictive performance of Reactron over existing product-only models by a large-scale reaction outcome prediction benchmark, and the adaptability of the model to learn new reactivity upon providing a few examples. Furthermore, it explores combinatorial reaction spaces, uncovering novel reactivities beyond its training data. With robust performance in both in- and out-of-distribution predictions, Reactron embodies human-like reasoning in chemistry and opens new frontiers in reaction discovery and synthesis design.