The findings suggest that modern Large Language Models can provide useful and consistent support for scientific peer review, however remaining differences between AI-generated and human-generated evaluations indicate that current systems should be viewed as complementary tools that assist human reviewers rather than replacements for expert human judgment.
AI-assisted peer review is increasingly discussed and adopted as a tool to support the scientific publishing process, yet there is little systematic understanding of how publication venues regulate its use or of how capable current AI review systems are. We address these questions by first surveying reviewer-facing AI policies across 111 leading AI/NLP conferences and medical journals, revealing substantial regulation differences between the two communities. Second, we evaluate AI-generated peer reviews at ICLR 2026 and Nature Communications using a novel dataset comprising original manuscript submissions and several hundred human- and machine-generated reviews. We compare reviews produced by open-source and proprietary models using complementary evaluation metrics, including LLM-as-a-Judge, score alignment, granularity, and overlap with human reviewers'concerns. Our results show that current LLMs can generate detailed and fluent reviews but exhibit systematic weaknesses, such as overly positive recommendations, generic criticism, and uneven evidence grounding. We demonstrate that aggregate quality scores alone can overestimate review quality and argue for multi-dimensional evaluation of AI-generated peer reviews.
Alexander M. Fichtl, Lukas Ellinger, Josefin Kelber et al.· 0 citations
Peer review is central to scientific decision-making, yet it is rarely evaluated or audited at scale. Growing submission volumes and the increasing use of large language models (LLMs) in drafting reviews have introduced new challenges for transparency, accountability, and quality control. Today, peer reviews are often produced through hybrid human--AI workflows, where a reviewer may develop the core evaluative ideas while using an LLM to refine wording, restructure arguments, or improve fluency. This shift raises new questions beyond authorship detection alone: Are reviews constructive? Are reviewer claims grounded in the submitted paper? How can we quantify collaboration between human reasoning and AI-assisted writing, and distinguish whether the intellectual contribution or the surface text originates from humans or models? In this industry talk, we present Reviewerly's retrieval-centered infrastructure for auditing peer review at scale. We describe three deployed systems: Peeriscope, which evaluates review quality across multiple interpretable dimensions; Peerispect, which uses retrieval-augmented generation (RAG) to verify whether reviewer claims are supported by evidence in the manuscript; and PeerPrism, which analyzes hybrid human--AI authorship by disentangling the origin of ideas from the origin of text in peer reviews, enabling measurement of how human reasoning and AI-generated writing interact within a review. We will share architectural design decisions, lessons learned from real-world deployment, and practical trade-offs among model complexity, interpretability, computational efficiency, and predictive accuracy. The talk will include short live demonstrations1 2 of our systems to illustrate how the combination of retrieval systems and LLM-based pipelines can improve peer review quality and strengthen transparency and research integrity across high-volume decision-making environments.
Negar Arabzadeh, Sajad Ebrahimi, Alireza Daghighfarsoodeh et al.· Annual International ACM SIG...· 0 citations
Large language models (LLMs) are increasingly used for research quality evaluation, with prior work exploring their scoring accuracy and the plausibility of review rationales. However, less is known about whether external score guidance changes the evidence presented in the generated review as well as the final score. This study uses 98 Allied Health Professions research outputs submitted for internal REF-style assessment, with specialist human review reports and adjudicated 1-4 reference scores. No-guidance baseline reviews are compared with oracle-guided reviews, where the supplied score is set to the rounded human reference score; extracted evaluation points are used to compare human and LLM evidence use. Using this design, oracle guidance improves scoring accuracy, with score-following checks showing that models do not simply copy the supplied score. Corrected score mismatches are associated with changes in the generated review frame, showing that the score signal can steer review rationales. This effect is direction-dependent: LLM reviews cover human strength or upgrade points more reliably than human weakness or downgrade points, with the weakest alignment for expert downgrade evidence. The results show that score-guided review generation can be evaluated at the level of review evidence, as well as the final score.
Background: The peer review system faces increasing strain from rising manuscript volumes, reviewer fatigue, and well-documented interreviewer disagreement. Large language models (LLMs) have shown potential to support the peer review process, but their ability to replicate editorial decisions at high-impact medical journals and their utility as manuscript screening tools remain unknown. Purpose: To compare the agreement between an LLM and the final editorial decision on manuscripts submitted to the American Journal of Sports Medicine and to evaluate the potential of LLMs as a manuscript screening tool. Study Design: Cross-sectional agreement study. Methods: Fifty-four manuscripts randomly selected from submissions to the American Journal of Sports Medicine (September 2024–October 2024) were reviewed by a locally deployed LLM (Ministral 3 14B; Mistral AI) using a standardized prompt. The artificial intelligence (AI) produced a categorical recommendation (reject, cascade, revision, or accept) and a numerical score (0-100) for each manuscript. Agreement with the final editorial decision was assessed by Cohen kappa (4-category model) for pooled human reviewers (n = 139 reviews) and the AI (n = 54). Screening performance was evaluated by positive predictive value (PPV), sensitivity, and specificity. Results: Pooled human reviewers demonstrated fair agreement with the final decision (κ = 0.181 [P < .001]; 42.4% agreement), while the AI demonstrated slight, nonsignificant agreement (κ = 0.126 [P = .099]; 37.0% agreement). The AI recommended revision for 61.1% of manuscripts, of which 72.7% were ultimately rejected or cascaded, demonstrating systematic “revision bias.” When the AI recommended rejection, 54.5% of those manuscripts were ultimately rejected and 27.3% were cascaded; when the AI recommended cascade, 50% were rejected and 50% were cascaded. However, when the AI recommended rejection or cascade (n = 21), 90.5% received a final decision of rejection or cascade (PPV, 90.5%; specificity, 81.8%). Manuscripts with an AI score <70 were rejected or cascaded 88.0% of the time (PPV, 88.0%). Conclusion: AI cannot replicate the nuanced judgment of human peer reviewers at a high-impact sports medicine journal. When AI recommended rejection or cascade, 90.5% of manuscripts received that final decision (descriptive PPV, 90.5%; 95% CI, 71.1%-97.3%), suggesting potential utility as an exploratory first-pass screening tool warranting further validation in larger cohorts. However, AI could not reliably distinguish manuscripts destined for outright rejection from those that would be cascaded to a sister journal—an important limitation for editorial triage applications.
Romir Patel, Christopher L. Shultz, Mathieu Ollivier et al.· American Journal of Sports M...· 0 citations
The results raise concerns that LLM-assisted evaluation may under-select proposals that human reviewers identify as highly novel, potentially reflecting the statistical logic of next-token prediction trained on past scientific outputs.
Background: Large language models (LLMs) are increasingly used to automate code review, but the reasoning behind their decisions remains hard to understand. Developers struggle to assess the validity of LLM-generated reviews, making it difficult to gauge how much trust to place in them. The role of Explainable AI (XAI) in code review and its impact on trust remain underexplored. Objective: We study the influence of XAI on developer trust in AI-assisted code reviews. Method: We conducted a within-subjects user study with 34 participants, comparing three LLM-based code review systems with varying levels of XAI support: Condition A (detailed explanation and review feedback), Condition B (review feedback only), and Condition C (no explanations). Participants reviewed real-world code change requests alongside the AI-generated reviews. We measured trust perceptions, agreement with the AI recommendation, the reasoning given for each decision, and the time taken. Results: The level of explanation significantly influences both trust and agreement with AI recommendations, but in different ways. Full explanations (A) yield the highest perceived trust (M = 3.99/5) but not the highest agreement, whereas moderate explanations (B) achieve the highest agreement (89.22%). This could suggest that more explanation prompts developers to question AI recommendations more frequently. No explanations (C) results in the lowest trust and agreement. Explanation level did not significantly affect review time. The most commonly cited reasons for decisions were code readability and correctness. Conclusion: Incorporating XAI into code review significantly changes trust perceptions and agreement with AI recommendations. These results inform the design and evaluation of trustworthy AI-based code review systems, as well as studies on the human factors of AI-assisted software development.
Zhenhan Gao, Marvin Muñoz Barón, Umm E. Habiba et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
