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

Hexuan Deng, Xiaopeng Ke, Yichen Li, Ruina Hu, Dehao Huang, Derek F. Wong, Yue Wang, Xuebo Liu, Min Zhang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:23997-24057, 2026.

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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-deng26j, title = {{C}o{C}o{R}eview{B}ench: A Completeness- and Correctness-Oriented Benchmark for {AI} Reviewers}, author = {Deng, Hexuan and Ke, Xiaopeng and Li, Yichen and Hu, Ruina and Huang, Dehao and Wong, Derek F. and Wang, Yue and Liu, Xuebo and Zhang, Min}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {23997--24057}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/deng26j/deng26j.pdf}, url = {https://proceedings.mlr.press/v306/deng26j.html}, 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.} }
Endnote
%0 Conference Paper %T CoCoReviewBench: A Completeness- and Correctness-Oriented Benchmark for AI Reviewers %A Hexuan Deng %A Xiaopeng Ke %A Yichen Li %A Ruina Hu %A Dehao Huang %A Derek F. Wong %A Yue Wang %A Xuebo Liu %A Min Zhang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-deng26j %I PMLR %P 23997--24057 %U https://proceedings.mlr.press/v306/deng26j.html %V 306 %X 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.
APA
Deng, H., Ke, X., Li, Y., Hu, R., Huang, D., Wong, D.F., Wang, Y., Liu, X. & Zhang, M.. (2026). CoCoReviewBench: A Completeness- and Correctness-Oriented Benchmark for AI Reviewers. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:23997-24057 Available from https://proceedings.mlr.press/v306/deng26j.html.

Related Material