MetaCaster: Meta-Harness-Optimized Agent for End-to-End Few-Shot Learning of Lightweight Time Series Forecasters
2026-08-24 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors address the challenge of making accurate time series forecasts when very little data is available, especially for small and specialized forecasting models. They propose MetaCaster, a system where multiple agents generate data and help train these small models using only a few examples and some text guidance. Instead of agents making forecasts directly, they prepare the forecasting models to be efficient and specific to the task. Experiments show that MetaCaster works well across many datasets and models, using less data and computing power while keeping good accuracy.
time series forecastingfew-shot learninglightweight forecastersmeta-learningagent-based systemsdata efficiencycomputational efficiencymulti-agent frameworksspecialized models
Authors
ChengAo Shen, Wenchao Yu, Fangyu Wu, Dongjin Song, Hanghang Tong, Dongsheng Luo, Wei Cheng, Haifeng Chen, Jingchao Ni
Abstract
Time series forecasting (TSF) is evolving toward multimodal and agentic settings, yet using foundation models remains uneconomical in resource-constrained scenarios, where compact, specialized forecasters are more desirable. However, lightweight forecasters typically require substantial training data, limiting their use in domains with scarce, slowly accumulated, or privacy-sensitive time series. To address this dilemma, we investigate the challenging problem of few-shot learning for lightweight forecasters. We propose MetaCaster, a meta-harness-optimized multi-agent framework that uses agentic data generation to automatically train specialized lightweight forecasters from only a few examples and textual contexts. Our work highlights a new TSF paradigm in which agents act not as forecasters but as intermediary engineers that prepare efficient, task-specific forecasters for deployment. Experiments on 18 datasets, 23 state-of-the-art lightweight forecasters, and 14 baselines demonstrate that MetaCaster achieves both data efficiency and computational efficiency while maintaining high-quality TSF performance.