The growing use of large language models (LLMs) in peer review threatens scholarly integrity.Recent conference policies allow AI tools for language polishing but prohibit their use for generating substantive content.However, existing detectors mainly rely on stylistic cues, making it difficult to distinguish between surface-level language refinement and genuine content generation.To address this, we advocate a content-based detection paradigm and introduce CoCoNUTS, a comprehensive benchmark containing 315,535 reviews covering leading AI conferences and six human-AI collaboration modes.Our evaluation shows that current detectors struggle to handle these nuanced settings.Consequently, we propose CoCoDet, an AI review detector designed to identify substantive AI-generation.Experiments demonstrate that CoCoDet achieves a macro F1-score of 98.24%.Crucially, on permissible machinepolished reviews, it maintains a low false positive rate of 3.89%, substantially outperforming the strongest baseline (7.84%).Examination on real-world reviews using CoCoDet reveals an escalating trend of substantive AI generation.Our work exposes the inadequacy of current detectors, underscoring the importance of domainspecific solutions.Our code is available at https://github.com/icip-cas/CoCoNUTS.

Dan Wang, Guozhao Mo, Yafei Shi, Cheng Zhang, Bo Zheng, Boxi Cao, Xuanang Chen, Yaojie Lu, Hongyu Lin, Ben He, Xianpei Han, Le Sun. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Jiawei Chen, Xinyan Guan, Qianhao Yuan, Mo Guozhao, Weixiang Zhou, Yaojie Lu, Hongyu Lin, Ben He, Le Sun, Xianpei Han. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

We propose a novel log-based anomaly detection method that leverages weakly supervised contrastive learning, named LogContrast. LogContrast aims to address the issues of limited and noisy labeled logs in real-world scenarios. During the training stage, LogContrast first augments the current log feature through dropout. Subsequently, it treats the current log feature and the augmented feature as a positive pair, while treating the current log feature and other log features in the same batch as negative pairs. The objective is to pull the positive pairs closer together and push the negative pairs farther apart, thereby encouraging similar logs to be closer to each other and dissimilar logs to be farther apart in the feature space. The experimental results demonstrate the excellent performance of LogContrast even with limited labeled logs and greater noise resistance compared to fully supervised methods. In addition, we explore the role of semantic features and demonstrate that semantic features have strong adaptability to constantly evolving logs.