arXiv 2506.17859

In-Context Learning Strategies Emerge Rationally

By Daniel Wurgaft, Ekdeep Singh Lubana, et al.

Published 2025-06-21

Mindmap

Browse the paper's core ideas, clusters, and relationships in a structured outline.

Recent work analyzing in-context learning (ICL) has identified a broad set of strategies that describe model behavior in different experimental conditions. We aim to unify these findings by asking why a model learns these disparate strategies in the first place. Specifically, we start with the observation that when trained to learn a mixture of tasks, as is popular in the literature, the strategies learned by a mode…

View the original paper on arXiv