Linear representation hypothesis: neural networks represent meaningful concepts as directions in their activation spaces.

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

• Superposition hypothesis: neural networks represent more features than dimensions using almost-orthogonal directions.hypothesis0.867
  Explanation for why dictionary learning can recover many more features than dimensions.
• "Representation geometry is a window into the inner world of neural networks."quote0.825
  The paper's concluding summary statement asserting the deep interpretive significance of representation geometry.
• Neural networks, trained with different objectives on different data and modalities, are converging to a shared statistical model of reality in their representation spaces.quote0.821
  The paper's central thesis statement, presented prominently after the abstract
• Smolensky (1986) proposes that viewing a neural representation under a basis that is not aligned with individual neurons can reveal the interpretable distributed structure of the neural representations.quote0.817
  Load-bearing theoretical claim providing the conceptual foundation for DAS.
• Neural networks show substantial alignment with biological representations in the brain, driven by shared task and data constraintsclaim0.815
  Extends convergence argument to brain-machine alignment
• Assuming linear representations enables identifying the location of certain variables in a DNN, but many insights fail to generalise when more powerful non-linear maps are usedclaim0.815
  Interpretive claim about what linear DAS results actually tell us
• Neural representations carry rich geometric structure; but does that structure causally shape behavior?quote0.814
  Opening sentence framing the paper's core inquiry.
• Neural Representations of Location Composed of Spatially Periodic Bands (Krupic et al., 2012)concept0.812
  Discovery of band cells; TEM-t also recapitulates these representations.