Nonparametric In-Context Learning under Growing Geometric Complexity: Minimax Optimality and Local Geometry-Adaptivity of Transformers

arXiv:2609.31458v1 Announce Type: new Abstract: Transformers have become a central architecture for in-context learning (ICL), particularly through their state-of-the-art performance in large language models. This success motivates understanding how transformers exploit task-relevant structure in g…

Read the full story at arXiv stat.ML ↗

Timeline · 1 report

1. 2026-09-28 04:00 · arXiv stat.ML
   Nonparametric In-Context Learning under Growing Geometric Complexity: Minimax Optimality and Local Geometry-Adaptivity of Transformers