Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMs
Abstract
Activation steering has emerged as a promising approach for efficiently adapting large language models (LLMs) to downstream behaviors. However, most existing steering methods rely on a single static direction per task or concept, making them inflexible under task variation and inadequate for complex tasks that require multiple coordinated capabilities. To address this limitation, we propose STEER2ADAPT, a lightweight framework that adapts LLMs by composing steering vectors rather than learning new ones from scratch. In many domains (e.g., reasoning or safety), tasks share a small set of underlying concept dimensions. STEER2ADAPT captures these dimensions as a reusable, low-dimensional semantic prior subspace, and adapts to new tasks by dynamically discovering a linear combination of basis vectors from only a handful of examples. Experiments across 9 tasks and 3 models in both reasoning and safety domains demonstrate the effectiveness of STEER2ADAPT, achieving an average improvement of 8.2%. Extensive analyses further show that STEER2ADAPT is a data-efficient, stable, and transparent inference-time adaptation method for LLMs.
Cite
@article{arxiv.2602.07276,
title = {Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMs},
author = {Pengrui Han and Xueqiang Xu and Keyang Xuan and Peiyang Song and Siru Ouyang and Runchu Tian and Yuqing Jiang and Cheng Qian and Pengcheng Jiang and Jiashuo Sun and Junxia Cui and Ming Zhong and Ge Liu and Jiawei Han and Jiaxuan You},
journal= {arXiv preprint arXiv:2602.07276},
year = {2026}
}