Synthetic In-Context Learning Task with Predefined Geometries

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

• Reasoning tasks with cyclic and sequential geometriesconcept0.783
  Language model experimental setting used to test manifold steering.
• The alignment between representation geometry and behavior geometry is not limited to days of the week but extends to months, letters, ages, and synthetic in-context learning tasks.claim0.764
  The paper's generalization claim, asserting that the days-of-week finding scales to other cyclic and structured concepts.
• Analogous alignment between representation manifold and behavior manifold is found across months, letters, ages, and synthetic in-context learning tasks in language models.finding0.759
  Generalization finding from the full paper extending beyond days-of-week to other structured concepts.
• In-Context Learning of Representations (Park et al. 2025)framework0.757
  Reports phase-like breakpoints and geometry changes as context scales; UCCT provides measurable predictor
• Universe Context for Synthetic Document Generationconcept0.754
  Comprehensive fictional background provided at each stage of synthetic document generation pipeline for consistency
• In-Context Learning as Optimizationconcept0.752
• manifold geometry principles extend to months, letters, ages, and in-context learning tasks across modalitiesfinding0.744
  Evidence that the weekday cyclic structure is not anomalous but reflects broader principle of concept geometry.
• Language models prefer reusing generic arithmetic mechanisms over learning task-specific modular computations even when task-specific geometry existsclaim0.741
  Broader interpretive claim about LM learning bias inferred from the findings