SOURCE-LINKED INTELLIGENCE

Enhancing SAE-based Steering via Neighbor Integrated Feature Selection

arXiv · AI, language, vision and robotics · article · Aug 28, 2026 · UTC

Sparse autoencoders (SAEs) disentangle model activations into interpretable features and are widely used for steering large language models. Most existing SAE-based steering methods select features by applying a top- filter based on statistical scores, assuming that higher-scoring features yield stronger steering effects. In this paper, we show that this assumption is often invalid, leading to suboptimal feature selection. Our analysis reveals that effective steering features may be distributed among representationally adjacent, semantically similar groups induced by feature splitting in SAEs.

Read original source ↗ Open in workspace

recordType

paper

region

Global