concept
active
concept:linear-representation-of-conceptsLinear Representation of Concepts
The established finding that transformer LLMs encode many interpretable concepts as linear directions in activation space
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (3)
concept
- Linear representationrelated_toThe idea that features are encoded as directions in activation space.
- Linear Representation of Featuresrelated_toThe central object of study — the idea that a concept like truth is encoded as a direction in the LLM's latent space
- The finding that interpretable concepts including character traits are encoded as linear directions in transformer residual streams
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.
- The hypothesis that models internalize concepts as approximately linear directions in representation space; used to interpret MDS injection behavior
- How a neural network encodes a semantic concept internally, argued to be better captured by manifolds than by atomic features.
- The sequential, continuous order of text, often challenged by diagrammatic branching.
- Correlative technique measuring the type of information encoded in distributed representations via linear predictability.
- Core contribution: the impasse where lifting linearity in alignment maps makes causal abstraction vacuous, but keeping it may miss non-linearly encoded features
- The idea that programs can be expressed as logical sentences, enabling direct deductive verification.
- Recent work identifying cases where LLM features are not one-dimensionally linear, a caveat to the linearity hypothesis.
- Hypothesis that information may be encoded in arbitrary non-linear subspaces of a neural network