High-Order Semantic Features

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• The features are often organized in geometrically-related clusters that share a semantic relationship.claim0.760
  Decoder cosine similarity maps onto concept similarity.
• Higher-Order Representationsconcept0.757
  Representations of one's own mental states; associated with consciousness in higher-order theories.
• semantic structureconcept0.755
  The meaningful organization of concepts in a model's representation space, claimed to be better captured by manifolds than by SAEs.
• The resulting features are highly abstract: multilingual, multimodal, and generalizing between concrete and abstract references.claim0.752
  Features respond to concepts across languages and in images, not just text.
• How can order relations in posets provide semantic models for data and computation?question0.741
  Motivating question throughout: using order theory to capture information flow, approximation, and program behavior.
• Higher-order Desiresconcept0.739
  Desires through which an agent identifies with some first-order desires and repudiates others.
• semantic anchoringconcept0.731
  The central idea that external structure binds latent patterns to desired targets.
• High Density Feature Clusterconcept0.728
  Mode in feature density histogram around 1e-5 corresponding to interpretable features, contrasted with ultralow density cluster