concept
active
concept:subspace-captureSubspace capture
A small fixed atom group spans the manifold subspace and the encoder consistently selects it.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Do Sparse Autoencoders Capture Concept Manifolds?introducesmentions
Frameworks (1)
framework
- Distinguishes direction-based (cone) and point-based (convex hull) dictionary learning.
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.
- Extension of DAS that learns a second rotation matrix on top of a fixed first one to decompose representations into sub-representations.
- The subspace of activation space spanned by the 171 orthogonalized emotion probe vectors, used to measure SAE feature emotional alignment
- Point-based SAE capture defined via Hausdorff distance of landmarks to the manifold.
- Intervention targeting specific dimensional subsets of activation vectors rather than full representations
- The multi-dimensional activation subspace whose directions causally mediate truthful behavior in LLMs
- Investigation of whether a distributed representation can be further decomposed into sub-representations encoding component identities.
- Ablates top-k PCs of character-count means vs random subspace.
- A vector subspace that causally impacts outputs only through the sign of its values, enabling harmless magnitude divergence