claim
active
claim:interventions-on-representations-causally-steer-belief-trajectories-with-effects-predictable-from-conceptual-geometryInterventions on representations causally steer belief trajectories with effects predictable from conceptual geometry
Finding (3).
Source paper
extracted_from(2026) · Eric Bigelow · Raphaël Sarfati · Daniel Wurgaft · Owen Lewis +4
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (4)
finding
- A linear model predicts out-of-sample steering effects on queried concepts from manifold distancessupportsPredictive geometry.
- Causal evidence.
- Hierarchy explains entanglement.
- Entanglement mirrors geometry.
Claims (1)
claim
- Central thesis.
Questions (1)
question
- Future work.
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.
- Core question motivating the shift from linear to geometry-aware steering; answered via manifold alignment analysis.
- Neural representation geometry causally shapes behavior; interventions respecting that geometry will yield natural trajectories.hypothesis0.822Central hypothesis tested via manifold steering experiments across language models and video world models.
- The use of interventions (rather than correlations) to establish a causal link between representation geometry and behavioral geometry.
- Open question.
- Central question: does geometry in activation space causally determine behavior?
- The central scientific question the paper addresses through the lens of interventional causality.
- The motivating research question of the paper
- Practical guidance for practitioners who lack ground-truth model organisms.