concept
active
concept:graph-structured-environmentsGraph-Structured Environments
Training environments formalized as graphs where nodes have sensory observations and edges represent actions; used to test structural generalization.
Neighborhood — ranked by edge-count
Methods (1)
method
- Training paradigm requiring prediction of upcoming sensory observations during spatial navigation across multiple environments sharing the same structure.
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.
- Training data with inherent geometric or relational structure, which induces geometric organization in model internals.
- A more complex geometric structure used to characterize in-context learning task representations
- Alexander's collaborative research center at UC Berkeley; included Sara Ishikawa and others whose contributions remain underrepresented in historical accounts.
- Graph structure constraining local interactions and ordering
- Tasks involving graph-structured geometries for in-context learning, used to test manifold steering.
- Core claim of the paper: the right level of description for neural representations is geometric structure mirroring the world.
- A structure created by an unfolding, differentiating process that adapts each part deeply, achieving mistake-free, complex, living geometry. Contrasted with fabricated structure.