license: apache-2.0
library_name: transformers
pipeline_tag: text-classification
ReviewClassify
This model is a component of CoCoReviewBench, a benchmark designed for the reliable and fine-grained evaluation of AI reviewers. Specifically, this model is used in the classification stage of the evaluation pipeline to categorize atomic opinions extracted from reviews into a predefined taxonomy.
- Paper: CoCoReviewBench: A Completeness- and Correctness-Oriented Benchmark for AI Reviewers
- Repository: GitHub - hexuandeng/CoCoReviewBench
Model Description
CoCoReviewBench addresses the challenge of evaluating AI-generated paper reviews. Since human reviews often cover only a subset of salient issues and sometimes contain mistakes, they are unreliable as gold references. To address this, the authors build category-specific benchmark subsets to strengthen Completeness and leverage reviewer-author-meta-review discussions to strengthen Correctness.
The ReviewClassify model is fine-tuned to classify discussion points into specific taxonomy labels, enabling category-level performance analysis of AI reviewers.
Usage
This model is intended for use within the CoCoReviewBench evaluation pipeline. For detailed instructions on generating AI reviews, splitting them into atomic opinions, and running the classification stage, please refer to the GitHub repository.
Citation
@inproceedings{deng2026cocoreviewbench,
title = {{CoCoReviewBench}: A Completeness- and Correctness-Oriented Benchmark for {AI} Reviewers},
author = {Deng, Hexuan and Li, Yichen and Ke, Xiaopeng and Hu, Ruina and Wong, Derek F. and Wang, Yue and Liu, Xuebo and Huang, Dehao and Zhang, Min},
booktitle = {Proceedings of the 43rd International Conference on Machine Learning},
series = {Proceedings of Machine Learning Research},
publisher = {PMLR},
year = {2026},
note = {To appear},
url = {https://github.com/hexuandeng/CoCoReviewBench}
}