CoCoReviewBench: A Completeness- and Correctness-Oriented Benchmark for AI Reviewers

2026-05-08

AI Summary

Current evaluations of AI peer review predominantly rely on surface-level agreement with human reviews, often overlooking factual correctness. However, human reviews themselves are frequently incomplete or erroneous, rendering them unreliable as a gold standard. This work proposes the first evaluation framework for AI peer review that jointly accounts for completeness and correctness. It constructs a high-quality subset of papers, leverages tripartite discussions among reviewers, authors, and meta-reviewers as expert annotations, and incorporates a dynamic filtering mechanism to exclude unreliable human reviews. Built on ICLR and NeurIPS data, the resulting CoCoReviewBench benchmark—comprising 3,900 papers—reveals that existing AI reviewers still exhibit significant deficiencies in factual accuracy and are prone to hallucination, while reasoning-based models demonstrate superior performance. This benchmark establishes a robust foundation and clear direction for future research in reliable AI-assisted peer review.

Abstract

Despite the rapid development of AI reviewers, evaluating such systems remains challenging: metrics favor overlap with human reviews over correctness. However, since human reviews often cover only a subset of salient issues and sometimes contain mistakes, they are unreliable as gold references. To address this, we build category-specific benchmark subsets and skip evaluation when the corresponding human reviews are missing to strengthen Completeness. We also leverage reviewer--author--meta-review discussions as expert annotations and filter unreliable reviews accordingly to strengthen Correctness. Finally, we introduce CoCoReviewBench, which curates 3,900 papers from ICLR and NeurIPS to enable reliable and fine-grained evaluation of AI reviewers. Analysis shows that AI reviewers remain limited in correctness and are prone to hallucinations, and highlights reasoning models as more effective reviewers, motivating further directions for improving AI reviewers. Benchmarks and models are available at https://github.com/hexuandeng/CoCoReviewBench.

Problem

• AI reviewers
• evaluation benchmark
• completeness
• correctness
• hallucination

Innovation

• AI Reviewers
• Benchmark
• Completeness
• Correctness
• Expert Annotations