Neural Representation Geometry

The broader conceptual framework that neural activations exhibit non-Euclidean geometric structure causally linked to behavior.

Neighborhood — ranked by edge-count

concept

neural behavior geometry
related_to
The manifold structure of model outputs, modelled by M_y.
Manifold Steering
implements
Central framework: steering neural networks by intervening along the curved manifold where a concept lives, rather than in straight lines through activation space.
activation manifold M_h
associated_with
Manifold fitted to representations in activation space.
Cyclic Geometry
associated_with
One of the geometric structures used to characterize reasoning task representations in language models, used as a test case for manifold steering
Graph Geometry
associated_with
A more complex geometric structure used to characterize in-context learning task representations
Physical Dynamics Geometry
associated_with
The geometry corresponding to physical dynamics, instantiated in the video world model test domain
Sequential Geometry
associated_with
A geometric structure characterizing sequential reasoning task representations, used as a test case for manifold steering

cosine ≥ 0.65 · no typed edge

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

Neural Geometry Frameworkframework0.840
Conceptual scheme introduced in this paper: neural networks develop internal geometric representations that mirror real-world geometry, providing the right level of description for interpretability and control.
Neural Geometriesconcept0.835
Neural representation geometry causally shapes behavior; interventions respecting that geometry will yield natural trajectories.hypothesis0.822
Central hypothesis tested via manifold steering experiments across language models and video world models.
"Representation geometry is a window into the inner world of neural networks."quote0.816
The paper's concluding summary statement asserting the deep interpretive significance of representation geometry.
Distributed Neural Representationsconcept0.803
Representations where individual neurons play multiple conceptual roles; patterns consisting of linear combinations of unit vectors.
There exists a bidirectional relationship between the geometry of neural representation and the geometry of model behaviorclaim0.801
Central empirical claim of the paper, demonstrated across tasks and modalities
Neural codeconcept0.794
The model's parameters considered as the actual 'code' implementing its algorithms, as opposed to human-written code.
Geometry unifies diverse neural architectures in machine learning systems.claim0.786