Does Joint-Embedding Predictive Architecture Pretraining Help Time Series Forecasting?
Abstract
Joint-embedding predictive architectures (JEPA) have emerged as a promising self-supervised pretraining paradigm for time series, learning representations by predicting target embeddings in latent space rather than reconstructing raw signals. Yet evidence on their benefits remains mixed, and most studies test only a single backbone or a narrow set of architectures, leaving unclear whether JEPA pretraining is a reliable improvement or one that depends heavily on the downstream model. We address this gap through a large scale evaluation of one JEPA instantiation across nine backbones and eleven benchmarks spanning temporal and spatio-temporal forecasting, the most extensive cross architecture assessment of JEPA for time series to date. We find that the benefit of this instantiation varies sharply across backbones, producing consistent gains for some architectures and consistent degradation for others, even on the same dataset. This pattern holds across both task families, indicating the variability is a general property of this instantiation rather than a dataset specific artifact worth accounting for when choosing a backbone in practice.
Related
- JEPADepth: Self-Supervised Monocular Depth Estimation via Joint-Embedding Predictive Architectures
- Local Nevanlinna embedding for Hardy space of Dirichlet series below critical exponent
- Informed forecasting: Leveraging auxiliary knowledge to boost large language model performance on time series forecasting
- Hierarchical Time Series Forecasting with Operator-Valued Kernels
- Series-wise Cross-validation for Global Time Series Forecasting
BibTeX
@misc{feng2026does,
title = {{Does Joint-Embedding Predictive Architecture Pretraining Help Time Series Forecasting?}},
author = {Yutong Feng and Bowen Liao and See Kiong Ng and Yuxuan Liang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.31680},
eprint = {2609.31680},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.31680},
}
RIS
TY - UNPB
TI - Does Joint-Embedding Predictive Architecture Pretraining Help Time Series Forecasting?
AU - Yutong Feng
AU - Bowen Liao
AU - See Kiong Ng
AU - Yuxuan Liang
PY - 2026
DA - 2026/09/16
PB - arXiv
DO - 10.48550/arxiv.2609.31680
UR - https://arxiv.org/abs/2609.31680
AB - Joint-embedding predictive architectures (JEPA) have emerged as a promising self-supervised pretraining paradigm for time series, learning representations by predicting target embeddings in latent space rather than reconstructing raw signals. Yet evidence on their benefits remains mixed, and most studies test only a single backbone or a narrow set of architectures, leaving unclear whether JEPA pretraining is a reliable improvement or one that depends heavily on the downstream model. We address this gap through a large scale evaluation of one JEPA instantiation across nine backbones and eleven benchmarks spanning temporal and spatio-temporal forecasting, the most extensive cross architecture assessment of JEPA for time series to date. We find that the benefit of this instantiation varies sharply across backbones, producing consistent gains for some architectures and consistent degradation for others, even on the same dataset. This pattern holds across both task families, indicating the variability is a general property of this instantiation rather than a dataset specific artifact worth accounting for when choosing a backbone in practice.
ER -