claim
active
claim:neural-networks-can-perform-computation-entirely-in-superposition-not-just-storageNeural networks can perform computation entirely in superposition, not just storage
Supported by abs model.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Toy Models of Superpositionmentions
Findings (2)
finding
- Computation in superposition.
- Mechanistic motif.
Questions (1)
question
- Open question.
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.
- Superposition hypothesis: neural networks represent more features than dimensions using almost-orthogonal directions.hypothesis0.845Explanation for why dictionary learning can recover many more features than dimensions.
- Theoretical model of how neural networks encode more features than dimensions, informing linear representation work.
- If we don't give a model enough neurons to implement this solution, how does it do computation in superposition?question0.834Left to future work.
- Interpretation of the cars-in-superposition circuit finding as an intentional representational strategy
- Superposition hypothesis: networks represent more features than neurons as almost-orthogonal directionshypothesis0.803Central hypothesis.
- The paper's central thesis statement, presented prominently after the abstract
- Extends convergence argument to brain-machine alignment