The spectral decoder successfully translates latent SAE interventions into physiologically interpretable frequency signatures such as slow-wave suppression and α-band restoration.

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

• Spectral decoder maps concept interventions to pathological slow-wave suppression and alpha-band restoration.finding0.886
  Physiological interpretability result linking latent steering to EEG frequency signatures.
• Spectral decoding of concept interventions can provide physiologically interpretable frequency signatures.claim0.844
  Claim that the spectral decoder adds physiological interpretability.
• A single SAE hyperparameter procedure driven by an intrinsic dictionary health audit transfers robustly across all three EEG transformer architectures.claim0.773
  Key methodological contribution claim about architecture-agnostic SAE tuning
• SAEs successfully extract sparse feature dictionaries from embeddings of SleepFM, REVE, and LaBraM EEG transformers.finding0.763
  Foundational empirical result enabling all downstream analysis
• We hypothesize that applying SAE-based mechanistic interpretability to EEG foundation models can expose representational failures and thereby improve clinical trust.hypothesis0.746
  Overarching motivating hypothesis of the paper
• SAE features can be grounded in clinical taxonomy (abnormality, age, sex, medication) to benchmark monosemanticity and entanglement.claim0.734
  Claim that feature grounding enables interpretability metrics.
• S suggests practical diagnostics for prompt design, retrieval, and light fine-tuning without additional training infrastructure.claim0.734
  Applied contribution.
• SAE training loss (MSE + L1 penalty with decoder norm scaling)method0.731
  The objective function combining L2 reconstruction error and L1 penalty scaled by decoder norm, used to train the SAE.