Latent Inference-Time Guidance of Time Series Foundation Models

arXiv:2609.38058 · stat.ML, cs.LG, stat.ME · Submitted 2026-09-29 · Read on arXiv

stat.ML, cs.LG, stat.ME

Submitted: 2026-09-29

Updated: 2026-09-29

Comments: 22 pages, 8 figures

Code: https://github.com/Dralliag/opera-python

License: http://creativecommons.org/licenses/by-nc-sa/4.0/

The gist: Time Series Foundation Models (TSFMs) currently provide state-of-the-art results in forecasting tasks.

Terminology

Abstract

Time Series Foundation Models (TSFMs) currently provide state-of-the-art results in forecasting tasks. They are available out-of-the-box and rely on in-context learning to make their predictions, which makes the quality of their performance highly sensitive to the user-selected lookback, covariates, horizon and training data distributions. In practise, the quality of the forecasts are variable but complementary, which highlights the need for a principled ensembling approach, rather than selecting the best context. This paper introduces Latent Inference-Time Guidance for TSFMs, which adaptively combines a pool of TSFM forecasts through a time-dependent latent space with independent components. The framework comes equipped with identifiability and reconstruction guarantees, whilst maintaining the off-the-shelf aspect of foundation models. We provide experiments on datasets at various frequencies and from multiple domains: these show that the approach is competitive with traditional ensembling approaches.

Sources

Related papers