@inproceedings{huanshuo-etal-2025-ctrla,
    title = "{C}trl{A}: Adaptive Retrieval-Augmented Generation via Inherent Control",
    author = "Huanshuo, Liu  and
      Zhang, Hao  and
      Guo, Zhijiang  and
      Wang, Jing  and
      Dong, Kuicai  and
      Li, Xiangyang  and
      Lee, Yi Quan  and
      Zhang, Cong  and
      Liu, Yong",
    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/landing_page/2025.findings-acl.652/",
    pages = "12592--12618",
    ISBN = "979-8-89176-256-5",
    abstract = "Retrieval-augmented generation (RAG) has emerged as a promising solution for mitigating hallucinations of large language models (LLMs) with retrieved external knowledge. Adaptive RAG enhances this approach by enabling dynamic retrieval during generation, activating retrieval only when the query exceeds LLM{'}s internal knowledge. Existing methods primarily focus on detecting LLM{'}s confidence via statistical uncertainty. Instead, we present the first attempts to solve adaptive RAG from a representation perspective and develop an inherent control-based framework, termed CtrlA. Specifically, we extract the features that represent the honesty and confidence directions of LLM and adopt them to control LLM behavior and guide retrieval timing decisions. We also design a simple yet effective query formulation strategy to support adaptive retrieval. Experiments show that CtrlA is superior to existing adaptive RAG methods on a diverse set of tasks. Honesty steering can effectively make LLMs more honest and confidence monitoring is a promising indicator of retrieval trigger."
}