Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Billa, Jayadev
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2604.15557
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911600865181696
author Billa, Jayadev
author_facet Billa, Jayadev
contents Steering vectors work for some concepts and layers but fail for others, and practitioners have no way to predict which setting applies before running an intervention. We introduce the Linear Accessibility Profile (LAP), a per-layer diagnostic that repurposes the logit lens as a predictor of steering vector effectiveness. The key measure, $A_{\mathrm{lin}}$, applies the model's unembedding matrix to intermediate hidden states, requiring no training. Across 24 controlled binary concept families on five models (Pythia-2.8B to Llama-8B), peak $A_{\mathrm{lin}}$ predicts steering effectiveness at $ρ= +0.86$ to $+0.91$ and layer selection at $ρ= +0.63$ to $+0.92$. A three-regime framework explains when difference-of-means steering works, when nonlinear methods are needed, and when no method can work. An entity-steering demo confirms the prediction end-to-end: steering at the LAP-recommended layer redirects completions on Gemma-2-2B and OLMo-2-1B-Instruct, while the middle layer (the standard heuristic) has no effect on either model.
format Preprint
id arxiv_https___arxiv_org_abs_2604_15557
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Predicting Where Steering Vectors Succeed
Billa, Jayadev
Machine Learning
Computation and Language
Steering vectors work for some concepts and layers but fail for others, and practitioners have no way to predict which setting applies before running an intervention. We introduce the Linear Accessibility Profile (LAP), a per-layer diagnostic that repurposes the logit lens as a predictor of steering vector effectiveness. The key measure, $A_{\mathrm{lin}}$, applies the model's unembedding matrix to intermediate hidden states, requiring no training. Across 24 controlled binary concept families on five models (Pythia-2.8B to Llama-8B), peak $A_{\mathrm{lin}}$ predicts steering effectiveness at $ρ= +0.86$ to $+0.91$ and layer selection at $ρ= +0.63$ to $+0.92$. A three-regime framework explains when difference-of-means steering works, when nonlinear methods are needed, and when no method can work. An entity-steering demo confirms the prediction end-to-end: steering at the LAP-recommended layer redirects completions on Gemma-2-2B and OLMo-2-1B-Instruct, while the middle layer (the standard heuristic) has no effect on either model.
title Predicting Where Steering Vectors Succeed
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2604.15557