Previous theoretical results pertaining to meta-learning on sequences build on contrived assumptions and are somewhat convoluted. We introduce new information-theoretic tools that lead to an elegant and very general decomposition of error into three components: irreducible error, meta-learning error, and intra-task error. These tools unify analyses across many meta-learning challenges. To illustrate, we apply them to establish new results about in-context learning with transformers. Our theoretical results characterizes how error decays in both the number of training sequences and sequence lengths. Our results are very general; for example, they avoid contrived mixing time assumptions made by all prior results that establish decay of error with sequence length.
@article{arxiv.2401.15530,
title = {An Information-Theoretic Analysis of In-Context Learning},
author = {Hong Jun Jeon and Jason D. Lee and Qi Lei and Benjamin Van Roy},
journal= {arXiv preprint arXiv:2401.15530},
year = {2024}
}