concept representation

How a neural network encodes a semantic concept internally, argued to be better captured by manifolds than by atomic features.

Neighborhood — ranked by edge-count

claim

Curved manifolds often represent concepts better than linear directions.
cites
Proposes that nonlinear geometric structure is superior to linear feature spaces for capturing semantic content.

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.

Conceptconcept0.843
Central entity of Jackson's framework: a structure invented to give coherent account of immediate consequences of actions; the building block of software design
Representation Designconcept0.825
The aspect of design dealing with data structures, modules, and implementation.
Concept Direction in Representation Spaceconcept0.820
A vector in activation space aligned with a behavioral concept; core object manipulated by RepE methods
Structure in representationsconcept0.817
The central question of whether representational geometry implies corresponding computational structure
concept geometryconcept0.812
The spatial/geometric organization of conceptual structure within neural network representations; central to the paper's thesis.
Concept (Formal Concept Analysis)concept0.805
Concept Algebraframework0.804
Probabilistic framework formalizing concept-specific subspaces for targeted steering in generative models.
concept vectorconcept0.798
Computed directional vector in activation space representing a specific concept, used for injection experiments