claim
active
claim:there-is-a-bidirectional-relationship-between-the-geometry-of-representation-and-behavior-across-tasks-and-modalitiesThere is a bidirectional relationship between the geometry of representation and behavior across tasks and modalities.
Author’s interpretive claim that the shared geometry is general and robust.
Source paper
extracted_from(2026) · Daniel Wurgaft · Can Rager · Matthew Kowal · Vasudev Shyam +12
Neighborhood — ranked by edge-count
Findings (2)
finding
- Central empirical result showing causal coupling between representation and behavior geometry across multiple substrates and modalities.
- Demonstrates bidirectional causal link: behavior manifold geometry can be recovered by optimizing in representation space.
Communities (3)
community
- Explores geometry of activation/behavior manifolds to enable selective, non-destructive concept interventions.
- Concepts encoded as curved manifolds and circular structures in LLM activation spaces.
- Neural network activation and behavior manifolds maintain geometric correspondence, enabling intervention optimization across language models and vision tasks.
Claims (1)
claim
- The paper's finding that the alignment holds in both directions — from representation to behavior and from behavior back to representation space.
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 finding that steering along M_h yields M_y behavior, and optimizing for M_y paths recovers M_h trajectories.
- Core finding: the structure models use internally (representations) is precisely reflected in their external behavior (outputs).
- Central empirical claim of the paper, demonstrated across tasks and modalities
- The paper's generalization claim, asserting that the days-of-week finding scales to other cyclic and structured concepts.
- The paper's deepest interpretive claim, asserting that representation structure and behavioral structure are not coincidentally aligned but deeply connected.
- The bidirectional correspondence M_h ↔ M_y, indicating that geometry in representation is not incidental but causally shapes behavior.
- How do representations differ or converge between architectures, tasks, and modalities?question0.848Broader research question MAS is positioned to address, citing multiple recent works.
Cross-corpus bridges (3)
same_concept_as · Nomic cosineExternal markdown files that talk about the same concept as this entity.
- aboutblank_kbWhat is the relationship between individual agency and collective biological organization, and how do lower-level components coordinate to produce higher-level goals?questions/what-is-the-relationship-between-individual-agency-and.md0.804
- aboutblank_kbWhat invariants and symmetries exist between biological processes operating in different problem spaces (metabolic, morphological, behavioral)?questions/what-invariants-and-symmetries-exist-between-biological-processes.md0.797
- aboutblank_kbHow do abstract mathematical and computational patterns (not determined by physics) get accessed and utilized by biological systems?questions/how-do-abstract-mathematical-and-computational-patterns-not.md0.785
Restated by (1)
cosine ≥ 0.90Other entities that say roughly the same thing. May be merge candidates or independent restatements across papers.