Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models
Abstract
Mechanistic Interpretability (MI) has emerged as a vital approach to demystify the opaque decision-making of Large Language Models (LLMs). However, existing reviews primarily treat MI as an observational science, summarizing analytical insights while lacking a systematic framework for actionable intervention. To bridge this gap, we present a practical survey structured around the pipeline: "Locate, Steer, and Improve." We formally categorize Localizing (diagnosis) and Steering (intervention) methods based on specific Interpretable Objects to establish a rigorous intervention protocol. Furthermore, we demonstrate how this framework enables tangible improvements in Alignment, Capability, and Efficiency, effectively operationalizing MI as an actionable methodology for model optimization. The curated paper list of this work is available at https://github.com/rattlesnakey/Awesome-Actionable-MI-Survey.
Keywords
Cite
@article{arxiv.2601.14004,
title = {Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models},
author = {Hengyuan Zhang and Zhihao Zhang and Mingyang Wang and Zunhai Su and Yiwei Wang and Qianli Wang and Shuzhou Yuan and Ercong Nie and Xufeng Duan and Feijiang Han and Qibo Xue and Zeping Yu and Chenming Shang and Xiao Liang and Jing Xiong and Hui Shen and Chaofan Tao and Zhengwu Liu and Senjie Jin and Zhiheng Xi and Dongdong Zhang and Sophia Ananiadou and Tao Gui and Ruobing Xie and Hayden Kwok-Hay So and Hinrich Schütze and Xuanjing Huang and Qi Zhang and Ngai Wong},
journal= {arXiv preprint arXiv:2601.14004},
year = {2026}
}