arxiv6.0 / 10

The Price of Choosing Examples: Context Overfitting in Adaptive In-Context Learning

Xiaoyu Li, Jiangxuan Long

Abstract

In-context learning is usually analyzed as if the examples in the prompt were sampled before the learner is chosen. In deployed systems they are not: examples are retrieved, filtered, re-ranked, edited, and repeatedly tested after seeing the query and the data store. We formalize this gap by treating prompt construction as adaptive data analysis through a conditional stochastic-kernel calculus. Our main quantity is the context capacity, the mutual information or approximate max-information between the data store and the final prompt transcript. For any frozen in-context learner and any adaptive context pipeline with capacity $κ$, we prove a finite-sample context generalization bound of order $\sqrt{κ/k}$, where $k$ is the number of selected examples whose losses form the reported score. We prove a matching lower bound, whose constant is exact when $\log k\llκ\ll k$, and derive two PAC-Bayesian retrieval rules: a first-order Gibbs rule that charges data-dependent candidate gates their prior mass, and a self-normalized second-order rule. On Banking77 with 64 retrieval recipes and two model families, the certificates hold in all 24,000 resampled selection events, and from 64 (Qwen2.5-7B) or 128 (GPT-4.1-mini) reused validation queries the Gibbs rule certifies a risk below the zero-shot error in over 99% of validation draws. Finally, we instantiate the theory for Bayesian linear in-context regression, including an order-statistics analysis of query-aligned top-$k$ Gaussian retrieval. The results separate context quality from context certification: longer or more relevant contexts improve risk only when the procedure used to choose them is stable enough for selected-context evidence to transfer.

Research area

retrieval-augmented generation (rag)robustnessstatistical learning theory
Published
16 Nov 2026
Source
arxiv
Org
University of New South Wales
View paper
Sign in to read and join the discussion.