See More, Forecast Better and Faster: Enhancing Time Series Foundation Models via Inference-Time Plug-and-Play Downsampling

Longlong Xu, Zeyan Li, Xiao He, Zhaoyang Yu, Dazhong Wen, Mingze Sun, Changhua Pei, Dan Pei
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:141296-141323, 2026.

Abstract

Time series foundation models (TSFMs) have demonstrated impressive generalization capabilities across diverse domains. However, they face significant challenges in long-term and ultra long-term forecasting. These challenges primarily arise from scalability limitations when TSFMs process extensive sequence lengths. To address this, we propose SPRINT, a training-free plug-and-play framework designed to empower TSFMs to see more, forecast better and faster during inference. The core idea is to perform forecasting in a downsampled-resolution space, enabling an extended look-back window with reduced computational costs. To avoid information loss and resolution mismatch caused by downsampling, SPRINT decomposes time series into trend and seasonal components, processing them separately. It predicts the low-frequency trend via a Resolution Interpolation workflow within the downsampled space, while preserving high-frequency details through a Pattern Replication mechanism for seasonality. Extensive experiments show that SPRINT achieves a significant improvement, increasing accuracy by 19% while enhancing efficiency with a reduction of max memory usage by 6.4$\times$ and inference time by 16.9$\times$ compared to state-of-the-art TSFMs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-xu26ac, title = {See More, Forecast Better and Faster: Enhancing Time Series Foundation Models via Inference-Time Plug-and-Play Downsampling}, author = {Xu, Longlong and Li, Zeyan and He, Xiao and Yu, Zhaoyang and Wen, Dazhong and Sun, Mingze and Pei, Changhua and Pei, Dan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {141296--141323}, 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/xu26ac/xu26ac.pdf}, url = {https://proceedings.mlr.press/v306/xu26ac.html}, abstract = {Time series foundation models (TSFMs) have demonstrated impressive generalization capabilities across diverse domains. However, they face significant challenges in long-term and ultra long-term forecasting. These challenges primarily arise from scalability limitations when TSFMs process extensive sequence lengths. To address this, we propose SPRINT, a training-free plug-and-play framework designed to empower TSFMs to see more, forecast better and faster during inference. The core idea is to perform forecasting in a downsampled-resolution space, enabling an extended look-back window with reduced computational costs. To avoid information loss and resolution mismatch caused by downsampling, SPRINT decomposes time series into trend and seasonal components, processing them separately. It predicts the low-frequency trend via a Resolution Interpolation workflow within the downsampled space, while preserving high-frequency details through a Pattern Replication mechanism for seasonality. Extensive experiments show that SPRINT achieves a significant improvement, increasing accuracy by 19% while enhancing efficiency with a reduction of max memory usage by 6.4$\times$ and inference time by 16.9$\times$ compared to state-of-the-art TSFMs.} }
Endnote
%0 Conference Paper %T See More, Forecast Better and Faster: Enhancing Time Series Foundation Models via Inference-Time Plug-and-Play Downsampling %A Longlong Xu %A Zeyan Li %A Xiao He %A Zhaoyang Yu %A Dazhong Wen %A Mingze Sun %A Changhua Pei %A Dan Pei %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-xu26ac %I PMLR %P 141296--141323 %U https://proceedings.mlr.press/v306/xu26ac.html %V 306 %X Time series foundation models (TSFMs) have demonstrated impressive generalization capabilities across diverse domains. However, they face significant challenges in long-term and ultra long-term forecasting. These challenges primarily arise from scalability limitations when TSFMs process extensive sequence lengths. To address this, we propose SPRINT, a training-free plug-and-play framework designed to empower TSFMs to see more, forecast better and faster during inference. The core idea is to perform forecasting in a downsampled-resolution space, enabling an extended look-back window with reduced computational costs. To avoid information loss and resolution mismatch caused by downsampling, SPRINT decomposes time series into trend and seasonal components, processing them separately. It predicts the low-frequency trend via a Resolution Interpolation workflow within the downsampled space, while preserving high-frequency details through a Pattern Replication mechanism for seasonality. Extensive experiments show that SPRINT achieves a significant improvement, increasing accuracy by 19% while enhancing efficiency with a reduction of max memory usage by 6.4$\times$ and inference time by 16.9$\times$ compared to state-of-the-art TSFMs.
APA
Xu, L., Li, Z., He, X., Yu, Z., Wen, D., Sun, M., Pei, C. & Pei, D.. (2026). See More, Forecast Better and Faster: Enhancing Time Series Foundation Models via Inference-Time Plug-and-Play Downsampling. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:141296-141323 Available from https://proceedings.mlr.press/v306/xu26ac.html.

Related Material