转向向量中的反向检测与控制

中文摘要 AI 辅助

转向向量（SVs）被广泛用于影响大型语言模型输出中概念（如真实性）的表达。SVs的一个关键假设是它们相对于该概念具有线性判别性：表现出该概念的文本表示与SV的对齐程度高于未表现出该概念的文本，这促使沿SV的正或负方向移动分别促进或抑制该概念。在本研究中，我们发现了一种反向检测-控制现象，即一些与正表示对齐的高判别性SVs会持续促进相反的行为，我们将这类向量称为反向转向向量（ISVs）。我们对ISVs的效果进行了几何表征，发现沿这些方向转向会系统性地将表示推向判别性下游头部，仿佛该概念不存在，甚至在解码前就已如此。基于此分析，我们提出了一种无需生成或相关响应评分即可区分ISVs的方法，这使得针对性的符号翻转成为可能，我们通过推理时干预（ITI）将其用于改进基于检测的基础转向流程。我们的方法在30项实验中的27项上取得了改进，提升幅度从+0.9%到+138%不等。我们在Gemma 3 12B、Qwen 2.5 14B和Olmo 3 7B这3个模型上，针对5个概念评估了我们的发现。

英文摘要

Steering vectors (SVs) are widely used to influence the expression of concepts (e.g., truthfulness) in large language model outputs. A key assumption underpinning SVs is that they are linearly discriminative with respect to the concept: representations of texts that exhibit the concept are more aligned with the SV than those that do not, motivating shifts along the positive or negative SV direction to respectively promote or suppress the concept. In this work, we identify an inverted detection-control phenomenon in which some highly discriminative SVs that are aligned with positive representations can consistently promote the opposite behavior. We refer to such vectors as inverted-steering vectors (ISVs). We provide a geometric characterization of ISVs' effects, finding that steering along these directions systematically pushes representations in discriminative downstream heads as if the concept were absent, even prior to decoding. Motivated by this analysis, we propose an approach for distinguishing ISVs without requiring generation or associated response scoring. This enables targeted sign flips, which we use to improve a foundational detection-based steering pipeline via Inference Time Intervention (ITI). Our approach improves results in 27/30 experiments, ranging from +0.9% to +138%. We evaluate our findings on Gemma 3 12B, Qwen 2.5 14B, and Olmo 3 7B across 5 concepts.

发表机构

• Northeastern University（东北大学）

机构由 AI 辅助整理，请以论文原文为准。