FlowState: Sampling-Rate-Equivariant Time-Series Forecasting

Lars Graf, Thomas Ortner, Stanisław Woźniak, Angeliki Pantazi
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:36584-36609, 2026.

Abstract

Existing time series foundation models (TSFMs), often based on transformer variants, lack adaptability to different sampling rates, struggle with generalization across varying context and target lengths and are computationally inefficient. We introduce FlowState, a novel TSFM architecture that achieves sampling-rate–equivariant forecasting through a unified design that pairs a state space model (SSM) encoder with a functional basis decoder (FBD). This design enables continuous-time modeling and dynamic time-scale adjustment, allowing FlowState to inherently generalize across all possible temporal resolutions, and dynamically adjust the forecasting horizons without retraining. We further propose an efficient pretraining strategy that improves robustness and accelerates training. Despite being one of the smallest TSFMs, FlowState achieves state-of-the-art results on the widely used GIFT-Eval benchmark, while demonstrating superior adaptability to unseen sampling rates. Our detailed analyses confirm the effectiveness of its components, and we demonstrate its unique ability to adapt to varying input sampling rates.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-graf26a, title = {{F}low{S}tate: Sampling-{R}ate-{E}quivariant {T}ime-{S}eries Forecasting}, author = {Graf, Lars and Ortner, Thomas and Wo\'{z}niak, Stanis{\l}aw and Pantazi, Angeliki}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {36584--36609}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/graf26a/graf26a.pdf}, url = {https://proceedings.mlr.press/v306/graf26a.html}, abstract = {Existing time series foundation models (TSFMs), often based on transformer variants, lack adaptability to different sampling rates, struggle with generalization across varying context and target lengths and are computationally inefficient. We introduce FlowState, a novel TSFM architecture that achieves sampling-rate–equivariant forecasting through a unified design that pairs a state space model (SSM) encoder with a functional basis decoder (FBD). This design enables continuous-time modeling and dynamic time-scale adjustment, allowing FlowState to inherently generalize across all possible temporal resolutions, and dynamically adjust the forecasting horizons without retraining. We further propose an efficient pretraining strategy that improves robustness and accelerates training. Despite being one of the smallest TSFMs, FlowState achieves state-of-the-art results on the widely used GIFT-Eval benchmark, while demonstrating superior adaptability to unseen sampling rates. Our detailed analyses confirm the effectiveness of its components, and we demonstrate its unique ability to adapt to varying input sampling rates.} }
Endnote
%0 Conference Paper %T FlowState: Sampling-Rate-Equivariant Time-Series Forecasting %A Lars Graf %A Thomas Ortner %A Stanisław Woźniak %A Angeliki Pantazi %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-graf26a %I PMLR %P 36584--36609 %U https://proceedings.mlr.press/v306/graf26a.html %V 306 %X Existing time series foundation models (TSFMs), often based on transformer variants, lack adaptability to different sampling rates, struggle with generalization across varying context and target lengths and are computationally inefficient. We introduce FlowState, a novel TSFM architecture that achieves sampling-rate–equivariant forecasting through a unified design that pairs a state space model (SSM) encoder with a functional basis decoder (FBD). This design enables continuous-time modeling and dynamic time-scale adjustment, allowing FlowState to inherently generalize across all possible temporal resolutions, and dynamically adjust the forecasting horizons without retraining. We further propose an efficient pretraining strategy that improves robustness and accelerates training. Despite being one of the smallest TSFMs, FlowState achieves state-of-the-art results on the widely used GIFT-Eval benchmark, while demonstrating superior adaptability to unseen sampling rates. Our detailed analyses confirm the effectiveness of its components, and we demonstrate its unique ability to adapt to varying input sampling rates.
APA
Graf, L., Ortner, T., Woźniak, S. & Pantazi, A.. (2026). FlowState: Sampling-Rate-Equivariant Time-Series Forecasting. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:36584-36609 Available from https://proceedings.mlr.press/v306/graf26a.html.

Related Material