claim

active

claim:geometry-arises-from-optimization-pressure-on-networks-trained-on-structured-data

Geometry arises from optimization pressure on networks trained on structured data.

Mechanistic explanation: geometric structure emerges naturally from standard training on data with underlying structure.

Source paper

extracted_from

The World Inside Neural Networks

(2026) · Geiger, Atticus · Lubana, Ekdeep Singh · Fel, Thomas · Merullo, Jack +3

Neighborhood — ranked by edge-count

Findings (1)

finding

In the Mountain Car case study, car position is a 1D manifold; linear interventions cross voids causing incoherence; following the 1D curve produces smooth control.
supports
Empirical demonstration that a semantically meaningful variable is encoded as a curved manifold, and that respecting its geometry is critical for effective intervention.

Communities (4)

community

Manifold-aware concept steering in neural representations
members_of
Explores geometry of activation/behavior manifolds to enable selective, non-destructive concept interventions.
Geometric concept representations in neural networks
members_of
Concepts encoded as curved manifolds and circular structures in LLM activation spaces.
Neural Geometry
cites
Neural geometry as fundamental computational substrate
members_of
Geometric structure in neural representations causally determines computation and behavior across diverse architectures, revealed through analysis of learned manifolds and cyclic concepts.

Concepts (2)

concept

structured data
cites
Training data with inherent geometric or relational structure, which induces geometric organization in model internals.
optimization pressure
cites
The force of gradient-based learning on structured data that drives networks to organize their representations into geometric structures.

Related by similarity (8)

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.

Networks compute on geometric manifolds and control should respect that geometry.claim0.802
Strong interpretive assertion linking discovery and control: neural computation is fundamentally manifold-structured.
Geometry unifies diverse neural architectures in machine learning systems.claim0.795
Networks encode structured geometric concepts that reflect external reality.claim0.782
Core claim of the paper: the right level of description for neural representations is geometric structure mirroring the world.
"Concept geometry provides a blueprint for controlling the behavior of neural networks—if you know how to look."quote0.771
The opening statement of the paper, framing concept geometry as the key to neural network control.
What seems like stark geometry also arises from unfolding.claim0.770
There is a bidirectional relationship between the geometry of representation and behavior across tasks and modalities.claim0.767
Author’s interpretive claim that the shared geometry is general and robust.
Neural representation geometry causally shapes behavior; interventions respecting that geometry will yield natural trajectories.hypothesis0.764
Central hypothesis tested via manifold steering experiments across language models and video world models.
In general, the geometry will be created by differentiation, not by addition or accretion; the parts are given their dimensions by differentiating operations within the spatial shell.claim0.758
A formal principle of the living process that uniqueness emerges from successive differentiation.