@inproceedings{long-etal-2025-llms,
    title = "{LLM}s Are Biased Towards Output Formats! Systematically Evaluating and Mitigating Output Format Bias of {LLM}s",
    author = "Long, Do Xuan  and
      Nguyen, Ngoc-Hai  and
      Sim, Tiviatis  and
      Dao, Hieu  and
      Joty, Shafiq  and
      Kawaguchi, Kenji  and
      Chen, Nancy F.  and
      Kan, Min-Yen",
    editor = "Chiruzzo, Luis  and
      Ritter, Alan  and
      Wang, Lu",
    booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
    month = apr,
    year = "2025",
    address = "Albuquerque, New Mexico",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/landing_page/2025.naacl-long.15/",
    pages = "299--330",
    ISBN = "979-8-89176-189-6",
    abstract = "We present the first systematic evaluation examining format bias in performance of large language models (LLMs). Our approach distinguishes between two categories of an evaluation metric under format constraints to reliably and accurately assess performance: one measures performance when format constraints are adhered to, while the other evaluates performance regardless of constraint adherence. We then define a metric for measuring the format bias of LLMs and establish effective strategies to reduce it. Subsequently, we present our empirical format bias evaluation spanning four commonly used categories{---}multiple-choice question-answer, wrapping, list, and mapping{---}covering 15 widely-used formats. Our evaluation on eight generation tasks uncovers significant format bias across state-of-the-art LLMs. We further discover that improving the format-instruction following capabilities of LLMs across formats potentially reduces format bias. Based on our evaluation findings, we study prompting and fine-tuning with synthesized format data techniques to mitigate format bias. Our methods successfully reduce the variance in ChatGPT{'}s performance among wrapping formats from 235.33 to 0.71 ({\%}{\textasciicircum}2)"
}