concept
active
concept:features-sae-featuresFeatures (SAE features)
Internal concepts a model uses, extracted as human-interpretable components.
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (2)
concept
- SAE featuressame_asThe individual, supposedly monosemantic directions learned by SAEs; argued here to fragment manifolds into disconnected pieces.
- Echo featuresassociated_withRepeated features activating on similar inputs, expected under L1 penalty.
Artifacts (1)
artifact
- preview.goodfire.aiaboutDesktop chat interface for feature understanding and steering; deprecated Feb 2026.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Method of adding scaled versions of sparse autoencoder latent features during generation to causally modulate model behavior
- Explains lower off-target effects of SAE methods.
- Interpretable concept representations hypothesized as basic units of computation.
- Interpretability tool underlying top-feature, details, steering tools.
- Domain of techniques for constructing informative features from raw data; covariance pooling is a feature engineering method for token sequences.
- Explains weaker datamix SAEs.
- Frequency-weighted spectral value πkλkj determining which features are retained.
- Backprop of logit-minus-mean-logit loss to surface causal intervention candidates, summed across token positions.