@inproceedings{li-etal-2025-curse,
    title = "Curse of Knowledge: Your Guidance and Provided Knowledge are biasing {LLM} Judges in Complex Evaluation",
    author = "Li, Weiyuan  and
      Wang, Xintao  and
      Yuan, Siyu  and
      Xu, Rui  and
      Chen, Jiangjie  and
      Dong, Qingqing  and
      Xiao, Yanghua  and
      Yang, Deqing",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.findings-emnlp.805/",
    doi = "10.18653/v1/2025.findings-emnlp.805",
    pages = "14900--14924",
    ISBN = "979-8-89176-335-7",
    abstract = "As large language models (LLMs) grow more capable, they face increasingly diverse and complex tasks, making reliable evaluation challenging. The paradigm of LLMs as judges has emerged as a scalable solution, yet prior work primarily focuses on simple settings. Their reliability in complex tasks{---}where multi-faceted rubrics, unstructured reference answers, and nuanced criteria are critical{---}remains understudied. In this paper, we constructed ComplexEval Bench, a challenge benchmark designed to systematically expose and quantify Auxiliary Information Induced Biases. We systematically investigated and validated 6 previously unexplored biases across 12 basic and 3 advanced scenarios. Key findings reveal: (1) all evaluated models exhibit significant susceptibility to these biases, with bias magnitude scaling with task complexity; (2) notably, Large Reasoning Models (LRMs) show paradoxical vulnerability. Our in-depth analysis offers crucial insights for improving the accuracy and verifiability of evaluation signals, paving the way for more general and robust evaluation models."
}