82% of features in 1M SAE had maximum Pearson correlation ≤0.3 with any MLP neuron, and manual inspection showed no semantic resemblance.

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

• Automated interpretability (Claude 3 Opus) and specificity scoring show SAE features are significantly more interpretable and specific than MLP neurons.finding0.840
  Quantitative comparison supporting SAE utility.
• Features in A/1 have median activation correlation of 0.72 with most similar feature in B/1; neurons have median 0.46finding0.807
  Systematic comparison showing features are substantially more universal than neurons across models
• Our SAEs' features are more interpretable than neurons.claim0.803
  Automated interpretability and specificity ratings show SAE features are clearer than MLP neurons.
• SAE features can be grounded in clinical taxonomy (abnormality, age, sex, medication) to benchmark monosemanticity and entanglement.claim0.786
  Claim that feature grounding enables interpretability metrics.
• Most correlated neuron A/neurons/470 has correlation of only 0.18 with base64 feature A/1/2357 and responds to code, HTML labels, URLsfinding0.785
  Shows base64 feature is polysemantic at neuron level but monosemantic as learned feature
• Monosemanticity and entanglement of SAE features were benchmarked for clinical taxonomy grounding across SleepFM, REVE, LaBraM.finding0.783
  Quantitative assessment of feature quality using clinical concepts across models.
• Text-based and self-steered emotionality ratings for SAE features are correlated at only ρ = +0.051 (n.s.).finding0.782
  Shows low agreement between the two evaluation modalities
• Spearman's rank correlation among different alignment metrics (CKA, SVCCA, Mutual k-NN, CKNNA) over 78 vision models is high across variants, with all p-values below 2.24×10^-105finding0.777
  Validates robustness of alignment metric choice