question
active
question:can-models-effectively-use-nonlinear-representationsCan models effectively use nonlinear representations?
Open question.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Toy Models of Superpositionmentions
Claims (1)
claim
- Intuition behind linear representations.
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.
- can we show that such graphs suitably represent the computation done by the model, and demonstrate their usefulness?question0.800Motivates case studies.
- Recent work identifying cases where LLM features are not one-dimensionally linear, a caveat to the linearity hypothesis.
- Interpretive claim about what linear DAS results actually tell us
- Selective pressure toward convergence via task generality
- Author's interpretation of the VTAB alignment results echoing Tolstoy
- The central scientific question the paper addresses through the lens of interventional causality.
- Open question A.1.1a.2.
- Hypothesis that information may be encoded in arbitrary non-linear subspaces of a neural network