Signatures of Steerability in Activation Space of Language Models

2026-09-12

AI Summary

研究通过对比表示来控制语言模型行为的有效性问题，使用简单的分离度量方法评估激活空间的可操纵性。

Abstract

Steering language models using a set of contrastive representations has been a canonical and computationally efficient method for controlling model behavior. Despite this success in controlling certain model behaviors, the effectiveness of activation steering varies markedly across concepts; the generalization properties of steering vectors are often considered a function of the dataset used to construct them. We make this dataset-dependence claim more rigorous and show that simple separation metrics strongly correlate with the downstream steerability of language models across diverse settings, even after controlling for layers and dataset effects. Beyond prediction, we provide evidence from a synthetic superposition experiment that separation metrics are strongly correlated with alignment between the empirical and true feature direction. Our results suggest that simple separability statistics can serve as practical diagnostics for when steering vectors are likely to work.

Problem

• steering
• language models
• activation space
• contrastive representations
• dataset dependence

Innovation

• separation metrics
• steerability
• language models