Median feature interval scored 12/14 on interpretability rubric vs median neuron score of 0

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

• Features in A/1 have median activation correlation of 0.72 with most similar feature in B/1; neurons have median 0.46finding0.780
  Systematic comparison showing features are substantially more universal than neurons across models
• 82% of features in 1M SAE had maximum Pearson correlation ≤0.3 with any MLP neuron, and manual inspection showed no semantic resemblance.finding0.760
  SAE features are not simply mirroring individual neurons.
• Higher-activating feature intervals are systematically more interpretable than lower-activating intervals in human analysisfinding0.747
  Shows interpretability correlates with activation strength, most model effect comes from high activations
• Feature Interpretability Rubricmethod0.740
  14-point scoring rubric for human evaluation of feature interpretability covering confidence, activation consistency, logit consistency, and specificity
• Automated interpretability (Claude 3 Opus) and specificity scoring show SAE features are significantly more interpretable and specific than MLP neurons.finding0.739
  Quantitative comparison supporting SAE utility.
• Arabic feature A/1/3450 has 27 neurons with coefficient magnitude ≥0.1 and three largest coefficients are negative; most correlated neuron responds to mixture of non-English languagesfinding0.737
  Demonstrates that the Arabic feature is not aligned to any single neuron
• Our SAEs' features are more interpretable than neurons.claim0.726
  Automated interpretability and specificity ratings show SAE features are clearer than MLP neurons.
• Most correlated neuron A/neurons/470 has correlation of only 0.18 with base64 feature A/1/2357 and responds to code, HTML labels, URLsfinding0.722
  Shows base64 feature is polysemantic at neuron level but monosemantic as learned feature