Read-Best Is Not Steer-Best: A Probing--Steering Layer Dissociation in Omni-Modal Large Language Models

Excerpt

arXiv:2609.22135v1 Announce Type: new Abstract: Omni-modal large language models integrate text, audio, and image signals into a shared residual stream, where concepts such as emotion can be linearly decoded and causally modified by activation steering. A common but rarely tested assumption is that the layer with the highest probing accuracy is also the best layer for steering, so injection layers are often selected by probe performance. We provide the first causal test of this assumption across