@inproceedings{hong-etal-2025-data,
    title = "Data Interpreter: An {LLM} Agent for Data Science",
    author = "Hong, Sirui  and
      Lin, Yizhang  and
      Liu, Bang  and
      Liu, Bangbang  and
      Wu, Binhao  and
      Zhang, Ceyao  and
      Li, Danyang  and
      Chen, Jiaqi  and
      Zhang, Jiayi  and
      Wang, Jinlin  and
      Zhang, Li  and
      Zhang, Lingyao  and
      Yang, Min  and
      Zhuge, Mingchen  and
      Guo, Taicheng  and
      Zhou, Tuo  and
      Tao, Wei  and
      Tang, Robert  and
      Lu, Xiangtao  and
      Zheng, Xiawu  and
      Liang, Xinbing  and
      Fei, Yaying  and
      Cheng, Yuheng  and
      Ni, Yongxin  and
      Gou, Zhibin  and
      Xu, Zongze  and
      Luo, Yuyu  and
      Wu, Chenglin",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/mtsummit-25-ingestion/2025.findings-acl.1016/",
    doi = "10.18653/v1/2025.findings-acl.1016",
    pages = "19796--19821",
    ISBN = "979-8-89176-256-5",
    abstract = "Large Language Model (LLM)-based agents have excelled in various domains but face significant challenges when applied to data science workflows due to their complex, multi-stage nature. Current LLM-based agents struggle with non-linear relationships, recursive dependencies, implicit data- and logic-dependent reasoning, and managing extensive context. In this paper, we introduce Data Interpreter, an LLM-based agent that addresses these challenges through hierarchical graph-based modeling to represent the complexity and a progressive strategy for step-by-step verification, refinement, and consistent context management. Extensive experiments confirm the effectiveness of Data Interpreter. On InfiAgent-DABench, it boosts performance by 25{\%} (from 75.9{\%} to 94.9{\%}), and on machine learning and open-ended tasks, it lifts accuracy from 88{\%} to 95{\%} and from 60{\%} to 97{\%}, respectively. Moreover, our method surpasses state-of-the-art baselines by 26{\%} on the MATH dataset. We will release the code upon publication."
}