concept
active
concept:dead-featuresDead features
SAE features that never activate on a large sample of data, indicating inefficient dictionary use.
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- Interpretable concept representations hypothesized as basic units of computation.
- Dual interpretation of features: in addition to responding to inputs, features also act to increase probability of specific output tokens
- The class of dead buildings, essentially all configurations except the living ones; almost coextensive with C_all.
- Extra feature representing unexplained error to make isomorphism exact.
- Fraction of training tokens on which a given feature has nonzero activation; used as proxy metric for autoencoder quality
- A feature that responds to only a single latent variable, contrasted with polysemantic features
- Premise that models represent concepts as features.
- Property that features activate on only a small fraction of inputs; enables compressed sensing and is what allows superposition to work