Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders

TL;DR

This paper investigates the internal representations of EEG foundation models using sparse autoencoders, revealing their interpretability, entanglement issues, and physiological relevance through spectral decoding.

Contribution

It introduces a framework applying TopK Sparse Autoencoders to interpret EEG models, benchmarking their representations, and analyzing their clinical and physiological implications.

Findings

Identified three operational regimes: steerable, entangled, and non-encoded concepts.

Exposed representational failures like 'wrecking-ball' interventions and confounding entanglements.

Mapped interventions to physiologically interpretable spectral signatures.

Abstract

EEG foundation models achieve state-of-the-art clinical performance, yet the internal computations driving their predictions remain opaque: a barrier to clinical trust. We apply TopK Sparse Autoencoders (SAEs) across three architecturally distinct EEG transformers: SleepFM, REVE, and LaBraM to extract sparse feature dictionaries from their embeddings. By grounding these features in a clinical taxonomy (abnormality, age, sex, and medication), we benchmark monosemanticity and entanglement across architectures. A single hyperparameter procedure, driven by an intrinsic dictionary health audit, transfers robustly across all three architectures. Via concept steering, we introduce a "target vs. off-target" probe area metric to quantify steering selectivity and reveal three operational regimes: selectively steerable, encoded but entangled, and non-encoded. This framework exposes critical representational failures: "wrecking-ball" interventions that collapse global model performance, and clinical entanglements, such as age-pathology confounding, where it is impossible to suppress one concept without corrupting the other. Finally, a spectral decoder maps these interventions back to the amplitude spectrum, translating latent manipulations into physiologically interpretable frequency signatures, such as pathological slow-wave suppression and $\alpha$-band restoration.