framework
active
framework:geometric-duality-of-sparse-conceptsGeometric Duality of Sparse Concepts
Distinguishes direction-based (cone) and point-based (convex hull) dictionary learning.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Do Sparse Autoencoders Capture Concept Manifolds?introducesmentions
Concepts (2)
concept
- Point-based SAE capture defined via Hausdorff distance of landmarks to the manifold.
- Subspace captureusesA small fixed atom group spans the manifold subspace and the encoder consistently selects it.
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.
- Any statement about a poset P yields a dual statement about P^∂ by interchanging ≤ and ≥; permits proof of one statement to establish its dual.
- Mechanistic finding by Bricken et al. 2023 about how LLMs store features; cited as operational justification for pattern-repository assumption
- Theoretical open question about the geometry of truth in LLMs raised in Discussion
- Central argumentative move of Section 3 separating hard problem from mapping problem.
- Discrete features and continuous manifolds are dual descriptions.
- No redundancy criterion.