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

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 .

AI reviewers are developing rapidly, but it remains challenging to evaluate how well they review papers. Existing evaluation methods often reward AI reviews for being similar to human reviews, rather than for being correct. However, human reviews often cover only some important issues and sometimes contain mistakes, so they are not always reliable as the only reference. To address this, we build evaluation subsets for different types of review issues and skip cases where the corresponding human reviews are missing, which helps evaluate completeness. We also use discussions among reviewers, authors, and meta-reviewers as expert annotations and filter out unreliable reviews, which helps evaluate correctness. Finally, we introduce CoCoReviewBench, which collects 3,900 papers from ICLR and NeurIPS to support reliable and fine-grained evaluation of AI reviewers. Our analysis shows that AI reviewers are still limited in correctness and are prone to hallucinations, while reasoning models are more effective reviewers. These findings suggest directions for improving AI reviewers. Benchmarks and models are available at .