Are the features extracted by SAEs from EEG transformers monosemantic or entangled?

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

• SAEs successfully extract sparse feature dictionaries from embeddings of SleepFM, REVE, and LaBraM EEG transformers.finding0.832
  Foundational empirical result enabling all downstream analysis
• Monosemanticity and entanglement of SAE features were benchmarked for clinical taxonomy grounding across SleepFM, REVE, LaBraM.finding0.804
  Quantitative assessment of feature quality using clinical concepts across models.
• SAE features can be grounded in clinical taxonomy (abnormality, age, sex, medication) to benchmark monosemanticity and entanglement.claim0.793
  Claim that feature grounding enables interpretability metrics.
• A single SAE hyperparameter procedure driven by an intrinsic dictionary health audit transfers robustly across all three EEG transformer architectures.claim0.777
  Key methodological contribution claim about architecture-agnostic SAE tuning
• Our SAEs' features are more interpretable than neurons.claim0.776
  Automated interpretability and specificity ratings show SAE features are clearer than MLP neurons.
• Sparse autoencoders extract features that are significantly more monosemantic than neurons, as shown by four independent lines of evidenceclaim0.767
  Central claim of the paper, supported by detailed feature analysis, human evaluation, automated interpretability of activations, and automated interpretability of logit weights
• We hypothesize that applying SAE-based mechanistic interpretability to EEG foundation models can expose representational failures and thereby improve clinical trust.hypothesis0.752
  Overarching motivating hypothesis of the paper
• EEG Transformer Embeddingsconcept0.743
  The internal representations of EEG transformers from which SAE features are extracted