@inproceedings{hao-etal-2026-balancesft,
    title = "{B}alance{SFT}: Improving {LLM} Function Calling with Balanced Training Signals and Data Hardness",
    author = "Hao, Bingguang  and
      Xu, Zengzhuang  and
      Wang, Maolin  and
      Wen, Yuntao  and
      Chen, Yicheng  and
      Peng, Cunyin  and
      Chen, Long  and
      Zhao, Xiangyu  and
      Gu, Jinjie  and
      Zhuang, Chenyi  and
      Zhang, Ji",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.900/",
    pages = "18094--18112",
    ISBN = "979-8-89176-395-1",
    abstract = "While Supervised Fine-Tuning (SFT) is the prevailing method for equipping Large Language Models (LLMs) with function calling capabilities, its effectiveness is often compromised by two critical challenges: 1) **Imbalanced Training Signals**, where lengthy Chain-of-Thought (CoT) reasoning tokens dominate the training signals over concise function calls in the learning objective, and 2) **Imbalanced Data Hardness**, characterized by a scarcity of hard training examples. To overcome these limitations, we propose Balanced Supervised Fine-tuning (**BalanceSFT**), a novel framework that incorporates two key components: a Self-adjusted Signal Balancing (SSB) loss that employs a learnable hyperparameter to dynamically adjust the token contributions of CoT reasoning and function calls, together with a Hard Data Re-sampling (HDR) strategy that establishes a feedback loop to selectively generate new, high-quality complex data guided by model errors. Extensive experiments demonstrate the effectiveness of our proposed BalanceSFT framework. With BalanceSFT, a 7B model achieves function calling performance that surpasses state-of-the-art models like GPT-5. Our code, models, and dataset are open-sourced."
}