concept
active
concept:superposition

Superposition

Phenomenon where models represent more features than dimensions via almost-orthogonal directions.

Neighborhood — ranked by edge-count

Thinkers (3)

thinker
  • Reported multimodal neurons in vision models including region neurons and person detector neurons
  • Earliest superposition reference in word embeddings.
  • Model-level superposition.

Frameworks (1)

framework
  • Compressed Sensing
    associated_with
    Mathematical framework enabling recovery of high-dimensional sparse vectors from low-dimensional projections; theoretically underpins sparse autoencoder approach

Methods (2)

method
  • Replacement model architecture with 30M features replacing MLP neurons.
  • Jacobian matching
    associated_with
    Penalizing difference between transcoder and MLP Jacobians, computed cheaply via random vector products.

Concepts (10)

concept
  • Weight superposition
    extendsrelated_to
    Virtual weights between features forced into superposition by low-dimensional projections.
  • Attentional features spread across multiple heads.
  • More computations than neurons/dimensions.
  • Polysemantic Neuron
    associated_with
    A neuron that responds to multiple unrelated inputs, posing a major challenge for circuit-level interpretation
  • Representation of features spread across multiple layers, complicating dictionary learning.
  • A recurring, abstract pattern found in circuits (e.g., equivariance, unioning over cases), inspired by circuit motifs in systems biology
  • Feature Sparsity
    associated_with
    Property that features activate on only a small fraction of inputs; enables compressed sensing and is what allows superposition to work
  • Cost of non-orthogonal feature embedding.
  • Privileged Basis
    associated_with
    A property of activations where neural network features align with basis dimensions due to sparse activation functions; absent in the residual stream but present in tokens, attention patterns, and MLP activations
  • Linear-model solution that trades off with superposition.

Questions (1)

question

Related by similarity (8)

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.