concept
active
concept:feature-as-neurons-in-sufficiently-large-modelsFeature (as neurons in sufficiently large models)
Working definition adopted by the paper.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Toy Models of Superpositionmentions
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.
- Core framing hypothesis; analogy not to be taken literally.
- First of three speculative claims forming the foundation of the circuits research agenda
- Scalar function of the input corresponding to a direction in the vector space of neuron activations; claimed to be the fundamental unit of neural networks
- Biologically plausible two-pool architecture from Krotov & Hopfield (2020) splitting self-attention into feature and memory neuron populations; used to interpret TEM-t place cells.
- Mirrors real networks.
- Clamping feature activations causally alters model behavior in interpretable ways.
- Features may not be strictly one-dimensional objects; higher-dimensional feature manifolds may exist in model representationshypothesis0.775Extension of superposition hypothesis to account for continuous families of features
- Data-induced competition over resources.