framework
active
framework:compressed-sensingCompressed Sensing
Mathematical framework enabling recovery of high-dimensional sparse vectors from low-dimensional projections; theoretically underpins sparse autoencoder approach
Neighborhood — ranked by edge-count
Papers (2)
paper
- Toy Models of Superpositionmentions
Concepts (1)
concept
- Superpositionassociated_withPhenomenon where models represent more features than dimensions via almost-orthogonal directions.
Frameworks (2)
framework
- Superposition Hypothesisassociated_withCore theoretical framework: neural networks represent more features than neurons by encoding features as directions in superposition
- Primary method introduced: trains a one-hidden-layer MLP with L1 sparsity penalty to decompose model activations into overcomplete feature dictionaries
Questions (2)
question
- Open 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.
- Maps transcoder features onto a library of ideal circuits via compressed sensing, with residual.
- The ongoing cost of maintaining counterfactual aspects of experience, conflating 'what is', 'what could be', 'what should be', and 'what will be'.
- Tanha reframed as the brain's compression drive pushing complexity toward simpler configurations.
- A process by which vasomotion collapses ambivalent neural patterns into durable definite states, reducing complexity.
- Method from Gurnee et al. 2023 for finding feature directions including individual neuron analysis
- Signal sent by fixed cells indicating the presence and location of stressed cells, enabling coordinated movement.
- Key capability: covariance pooling compresses gigabytes of activations into compact stable embeddings without large labeled datasets.
- Attention probe predicting whether an Entity is hallucinated; AUC .94 on validation.