@inproceedings{yao-etal-2025-gradot,
    title = "{G}rad{OT}: Training-free Gradient-preserving Offsite-tuning for Large Language Models",
    author = "Yao, Kai  and
      Tan, Zhaorui  and
      Gao, Penglei  and
      Li, Lichun  and
      Wu, Kaixin  and
      Wang, Yinggui  and
      Zhao, Yuan  and
      Ji, Yixin  and
      Zhu, Jianke  and
      Wang, Wei",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/landing_page/2025.acl-long.255/",
    pages = "5115--5130",
    ISBN = "979-8-89176-251-0",
    abstract = "The rapid growth of large language models (LLMs) with traditional centralized fine-tuning emerges as a key technique for adapting these models to domain-specific challenges, yielding privacy risks for both model and data owners. One promising solution, called offsite-tuning (OT), is proposed to address these challenges, where a weaker emulator is compressed from the original model and further fine-tuned with adapter to enhance privacy. However, the existing OT-based methods require high computational costs and lack theoretical analysis. This paper introduces a novel OT approach based on gradient-preserving compression. By analyzing the OT problem through the lens of optimization, we propose a method that selectively applies compression techniques such as rank compression and channel pruning, preserving the gradients of fine-tuned adapters while ensuring privacy. Extensive experiments demonstrate that our approach surpasses existing OT methods, both in terms of privacy protection and model performance. Our method provides a theoretical foundation for OT and offers a practical, training-free solution for offsite-tuning of large-scale LLMs."
}