Age and pathology are clinically entangled in EEG foundation model representations such that suppressing one concept inevitably corrupts the other.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• EEG foundation models achieve state-of-the-art clinical performance yet their internal computations remain opaque, constituting a barrier to clinical trust.claim0.820
  Motivating claim for the entire paper
• Age-pathology confounding observed: impossible to suppress one concept without corrupting the other.finding0.819
  Empirical demonstration of entanglement between age and pathology features.
• Clinical concepts in EEG foundation models fall into three operational regimes: selectively steerable, encoded but entangled, and non-encoded.claim0.817
  Interpretive claim summarizing the spectrum of concept steerability discovered.
• age-pathology confounding, where it is impossible to suppress one concept without corrupting the otherquote0.813
  Verbatim description of age-pathology entanglement.
• Age-pathology confounding prevents independent steering of age and pathology concepts.claim0.808
  Interpretive assertion about clinical entanglement in the representations.
• We hypothesize that applying SAE-based mechanistic interpretability to EEG foundation models can expose representational failures and thereby improve clinical trust.hypothesis0.785
  Overarching motivating hypothesis of the paper
• What clinical concepts are encoded in the internal representations of EEG foundation models?question0.780
  Primary research question driving the extraction and benchmarking of SAE features
• SAE features can be grounded in clinical taxonomy (abnormality, age, sex, medication) to benchmark monosemanticity and entanglement.claim0.775
  Claim that feature grounding enables interpretability metrics.