English

Technical Debt in In-Context Learning: Diminishing Efficiency in Long Context

Machine Learning 2025-10-28 v2 Artificial Intelligence

Abstract

Transformers have demonstrated remarkable in-context learning (ICL) capabilities, adapting to new tasks by simply conditioning on demonstrations without parameter updates. Compelling empirical and theoretical evidence suggests that ICL, as a general-purpose learner, could outperform task-specific models. However, it remains unclear to what extent the transformers optimally learn in-context compared to principled learning algorithms. To investigate this, we employ a meta ICL framework in which each prompt defines a distinctive regression task whose target function is drawn from a hierarchical distribution, requiring inference over both the latent model class and task-specific parameters. Within this setup, we benchmark sample complexity of ICL against principled learning algorithms, including the Bayes optimal estimator, under varying performance requirements. Our findings reveal a striking dichotomy: while ICL initially matches the efficiency of a Bayes optimal estimator, its efficiency significantly deteriorates in long context. Through an information-theoretic analysis, we show that the diminishing efficiency is inherent to ICL. These results clarify the trade-offs in adopting ICL as a universal problem solver, motivating a new generation of on-the-fly adaptive methods without the diminishing efficiency.

Keywords

Cite

@article{arxiv.2502.04580,
  title  = {Technical Debt in In-Context Learning: Diminishing Efficiency in Long Context},
  author = {Taejong Joo and Diego Klabjan},
  journal= {arXiv preprint arXiv:2502.04580},
  year   = {2025}
}

Comments

NeurIPS 2025 Camera Ready

R2 v1 2026-06-28T21:35:35.946Z