Mechanistic Indicators of Steering Effectiveness in Large Language Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866917334337191936 |
|---|---|
| author | Jafari, Mehdi Xue, Hao Salim, Flora |
| author_facet | Jafari, Mehdi Xue, Hao Salim, Flora |
| contents | Activation-based steering enables Large Language Models (LLMs) to exhibit targeted behaviors by intervening on intermediate activations without retraining. Despite its widespread use, the mechanistic factors that govern when steering succeeds or fails remain poorly understood, as prior work has relied primarily on black-box outputs or LLM-based judges. In this study, we investigate whether the reliability of steering can be diagnosed using internal model signals. We focus on two information-theoretic measures: the entropy-derived Normalized Branching Factor (NBF), and the Kullback-Leibler (KL) divergence between steered activations and targeted concepts in the vocabulary space. We hypothesize that effective steering corresponds to structured entropy preservation and coherent KL alignment across decoding steps. Building on a reliability study demonstrating high inter-judge agreement between two architecturally distinct LLMs, we use LLM-generated annotations as ground truth and show that these mechanistic signals provide meaningful predictive power for identifying successful steering and estimating failure probability. We further introduce a stronger evaluation baseline for Contrastive Activation Addition (CAA) and Sparse Autoencoder-based steering, the two most widely adopted activation-steering methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01716 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Mechanistic Indicators of Steering Effectiveness in Large Language Models Jafari, Mehdi Xue, Hao Salim, Flora Computation and Language Activation-based steering enables Large Language Models (LLMs) to exhibit targeted behaviors by intervening on intermediate activations without retraining. Despite its widespread use, the mechanistic factors that govern when steering succeeds or fails remain poorly understood, as prior work has relied primarily on black-box outputs or LLM-based judges. In this study, we investigate whether the reliability of steering can be diagnosed using internal model signals. We focus on two information-theoretic measures: the entropy-derived Normalized Branching Factor (NBF), and the Kullback-Leibler (KL) divergence between steered activations and targeted concepts in the vocabulary space. We hypothesize that effective steering corresponds to structured entropy preservation and coherent KL alignment across decoding steps. Building on a reliability study demonstrating high inter-judge agreement between two architecturally distinct LLMs, we use LLM-generated annotations as ground truth and show that these mechanistic signals provide meaningful predictive power for identifying successful steering and estimating failure probability. We further introduce a stronger evaluation baseline for Contrastive Activation Addition (CAA) and Sparse Autoencoder-based steering, the two most widely adopted activation-steering methods. |
| title | Mechanistic Indicators of Steering Effectiveness in Large Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2602.01716 |