Non-Linear Representation Hypothesis

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

• Non-Linear Representation Dilemmaconcept0.824
  Core contribution: the impasse where lifting linearity in alignment maps makes causal abstraction vacuous, but keeping it may miss non-linearly encoded features
• Linear representationconcept0.803
  The idea that features are encoded as directions in activation space.
• Non-Linear Representations in LLMsconcept0.789
  Recent work identifying cases where LLM features are not one-dimensionally linear, a caveat to the linearity hypothesis.
• Linear representation hypothesis: neural networks represent meaningful concepts as directions in their activation spaces.hypothesis0.785
  Foundation for interpreting features as linear directions.
• 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 usedclaim0.779
  Interpretive claim about what linear DAS results actually tell us
• Results collectively provide strong evidence that some version of the superposition hypothesis and linear representation hypothesis is trueclaim0.765
  Authors' overall conclusion from number of interpretable features, activation-level correspondence to intensity, sensible logit weights, and interference weights
• Linear Representation of Featuresconcept0.761
  The central object of study — the idea that a concept like truth is encoded as a direction in the LLM's latent space
• How can non-linear reflection dynamics be formalized using probabilistic modeling and information theory?question0.749
  Theoretical open question about the mathematical treatment of reflection mechanisms.