arXiv：cs.LG（机器学习，全量分类）· Laziz U. Abdullaev, Minh-Hieu Pham, Bach Do, Khoat Than, Tan M. Nguyen·· 14 小时前AI 评分36
Kernelized Activation Steering：将激活引导提升到再生核希尔伯特空间
Kernelized Activation Steering
AI 导读
研究者提出 Kernelized Activation Steering（KAS），将激活引导建模为纯核函数评估的优化问题，无需构建显式特征映射即可生成依赖激活值的隐式引导分数。KAS 实现局部自适应引导，DiM 在线性核下成为其特例，更丰富的核可带来几何感知的干预。在越狱 LLM 和图像风格控制等标准激活引导任务上，KAS 表现优于或持平现有方法。
正文
Abstract:Activation steering provides a simple, training-free mechanism for controlling attributes of generative models such as sentiment, style, and helpfulness. However, standard approaches such as Difference-in-Means apply a single input-independent steering vector across all activations, limiting expressivity and ignoring the local geometry of the activation space. We propose Kernelized Activation Steering (KAS), a unifying framework that lifts activation steering into a reproducing kernel Hilbert space. KAS formulates steering as an optimization problem expressed purely via kernel evaluations, yielding an implicit, activation-dependent steering score without constructing explicit feature maps. Unlike DiM, KAS induces locally adaptive steering: each activation is modified according to its relative position with respect to source and target reference sets, producing a nonlinear steering field over the representation space. Importantly, DiM is recovered as a special case under a linear kernel, while richer kernels enable geometry-aware interventions. Across standard activation steering tasks, including jailbreaking LLMs and image style control, KAS outperforms or is on par with the existing methods.
| Comments: | NeurIPS 2026 |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.01062 [cs.LG] |
| (or arXiv:2610.01062v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.01062 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Laziz Abdullaev [view email]
[v1]
Thu, 1 Oct 2026 05:00:53 UTC (6,687 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org