Action Features

Dual interpretation of features: in addition to responding to inputs, features also act to increase probability of specific output tokens

Neighborhood — ranked by edge-count

concept

Token-in-Context Feature
associated_with
Feature that fires on a specific token only within a specific surrounding context (e.g., 'the' in physics vs 'the' in mathematics)

cosine ≥ 0.65 · no typed edge

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

actionconcept0.839
Changing configuration to sample environment differently; minimizes free energy.
Action Selectionconcept0.809
Choice of policies minimizing expected free energy to realize preferred future states.
feature as applicationconcept0.796
Metaphor treating each system feature or function as a separate application that can be independently loaded and managed.
Feature engineeringconcept0.760
Domain of techniques for constructing informative features from raw data; covariance pooling is a feature engineering method for token sequences.
Pure Featureconcept0.757
A feature that responds to only a single latent variable, contrasted with polysemantic features
Feature Sparsityconcept0.754
Property that features activate on only a small fraction of inputs; enables compressed sensing and is what allows superposition to work
Dead featuresconcept0.752
SAE features that never activate on a large sample of data, indicating inefficient dictionary use.
Context Featureconcept0.751
Feature that activates across all tokens within a specific context (e.g., DNA sequences, base64 strings)