question
active
question:are-concept-likelihood-functions-implemented-in-a-non-linear-way-and-are-they-implemented-in-earlier-or-later-layers-relative-to-this-linear-belief-representationare concept likelihood functions implemented in a non-linear way, and are they implemented in earlier or later layers relative to this linear belief representation?
Open question.
Source paper
extracted_from(2025) · Eric Bigelow · Daniel Wurgaft · YingQiao Wang · Noah D. Goodman +3
Neighborhood — ranked by edge-count
Papers (1)
paper
- Belief Dynamics Reveal the Dual Nature of In-Context Learning and Activation Steeringassociated_withmentions
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.
- Open question.
- Quantitative relationship between concept frequency and feature presence.
- Speculative interpretation of breakdown.
- Interpretive claim about what linear DAS results actually tell us
- Establishes that the observed linear structure is not merely a representation of text probability
- How do the theory and model generalize to non-binary concept spaces with multiple belief directions?question0.762Future work.
- Hypothesis that information may be encoded in arbitrary non-linear subspaces of a neural network
- Linear representation hypothesis: neural networks represent meaningful concepts as directions in their activation spaces.hypothesis0.752Foundation for interpreting features as linear directions.