English

In-context Time Series Predictor

Machine Learning 2026-02-06 v2 Artificial Intelligence Computation and Language Machine Learning

Abstract

Recent Transformer-based large language models (LLMs) demonstrate in-context learning ability to perform various functions based solely on the provided context, without updating model parameters. To fully utilize the in-context capabilities in time series forecasting (TSF) problems, unlike previous Transformer-based or LLM-based time series forecasting methods, we reformulate "time series forecasting tasks" as input tokens by constructing a series of (lookback, future) pairs within the tokens. This method aligns more closely with the inherent in-context mechanisms, and is more parameter-efficient without the need of using pre-trained LLM parameters. Furthermore, it addresses issues such as overfitting in existing Transformer-based TSF models, consistently achieving better performance across full-data, few-shot, and zero-shot settings compared to previous architectures.

Keywords

Cite

@article{arxiv.2405.14982,
  title  = {In-context Time Series Predictor},
  author = {Jiecheng Lu and Yan Sun and Shihao Yang},
  journal= {arXiv preprint arXiv:2405.14982},
  year   = {2026}
}

Comments

Camera-ready version. Accepted at ICLR 2025

R2 v1 2026-06-28T16:37:58.108Z