Test-Time Efficient Pretrained Model Portfolios for Time Series Forecasting
Abstract
Is bigger always better for time series foundation models? With the question in mind, we explore an alternative to training a single, large monolithic model: building a portfolio of smaller, pretrained forecasting models. By applying ensembling or model selection over these portfolios, we achieve competitive performance on large-scale benchmarks using much fewer parameters. We explore strategies for designing such portfolios and find that collections of specialist models consistently outperform portfolios of independently trained generalists. Remarkably, we demonstrate that post-training a base model is a compute-effective approach for creating sufficiently diverse specialists, and provide evidences that ensembling and model selection are more compute-efficient than test-time fine-tuning.
Cite
Text
Kayaalp et al. "Test-Time Efficient Pretrained Model Portfolios for Time Series Forecasting." International Conference on Learning Representations, 2026.Markdown
[Kayaalp et al. "Test-Time Efficient Pretrained Model Portfolios for Time Series Forecasting." International Conference on Learning Representations, 2026.](https://mlanthology.org/iclr/2026/kayaalp2026iclr-testtime/)BibTeX
@inproceedings{kayaalp2026iclr-testtime,
title = {{Test-Time Efficient Pretrained Model Portfolios for Time Series Forecasting}},
author = {Kayaalp, Mert and Turkmen, Ali Caner and Shchur, Oleksandr and Mercado, Pedro and Ansari, Abdul Fatir and Bohlke-Schneider, Michael and Wang, Bernie},
booktitle = {International Conference on Learning Representations},
year = {2026},
url = {https://mlanthology.org/iclr/2026/kayaalp2026iclr-testtime/}
}