Analyze Feature Flow to Enhance Interpretation and Steering in Language Models

TL;DR

This paper presents a novel, data-free method to trace feature evolution across layers in large language models, enhancing interpretability and enabling targeted control over model outputs.

Contribution

It introduces a cosine similarity-based approach to map feature flow across layers, allowing for detailed interpretability and direct steering of language models.

Findings

Granular feature flow graphs reveal feature persistence and transformation.

Cross-layer feature maps enable targeted manipulation of model behavior.

Method improves understanding and control of large language models.

Abstract

We introduce a new approach to systematically map features discovered by sparse autoencoder across consecutive layers of large language models, extending earlier work that examined inter-layer feature links. By using a data-free cosine similarity technique, we trace how specific features persist, transform, or first appear at each stage. This method yields granular flow graphs of feature evolution, enabling fine-grained interpretability and mechanistic insights into model computations. Crucially, we demonstrate how these cross-layer feature maps facilitate direct steering of model behavior by amplifying or suppressing chosen features, achieving targeted thematic control in text generation. Together, our findings highlight the utility of a causal, cross-layer interpretability framework that not only clarifies how features develop through forward passes but also provides new means for transparent manipulation of large language models.