mlsanityData to 7 Oct 2026

Lightweight Online Adaption for Time Series Foundation Model Forecasts

arXiv 2502.12920 ↗First posted 18 Feb 2025cs.LG stat.MLcited by 14

Not on the map. Why?

Abstract (arXiv)

Foundation models (FMs) have emerged as a promising approach for time series forecasting. While effective, FMs typically remain fixed during deployment due to the high computational costs of learning them online. Consequently, deployed FMs fail to adapt their forecasts to current data characteristics, despite the availability of online feedback from newly arriving data. This raises the question of whether FM performance can be enhanced by the efficient usage of this feedback. We propose ELF to answer this question. ELF is a lightweight mechanism for the online adaption of FM forecasts in response to online feedback. ELF consists of two parts: a) the ELF-Forecaster which is used to learn the current data distribution; and b) the ELF-Weighter which is used to combine the forecasts of the FM and the ELF-Forecaster. We evaluate the performance of ELF in conjunction with several recent FMs across a suite of standard time series datasets. In all of our experiments we find that using ELF improves performance. This work demonstrates how efficient usage of online feedback can be used to improve FM forecasts.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 39 of its 66 references that are in the map. First posted on arXiv 18 Feb 2025.

Builds on

39 of 66 references.

References · 39

  1. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  2. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,339
  3. 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
  4. 2021Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecastingcited by 1,654
  5. 2018An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modelingcited by 1,447
  6. 2023A Comprehensive Survey of Continual Learning: Theory, Method and Applicationcited by 860
  7. 2024Chronos: Learning the Language of Time Seriescited by 712
  8. 2023A decoder-only foundation model for time-series forecastingcited by 650
  9. 2024Unified Training of Universal Time Series Forecasting Transformerscited by 588
  10. 2020Time Series Forecasting With Deep Learning: A Surveycited by 457
  11. 2021Monash Time Series Forecasting Archivecited by 256
  12. 2024Foundation Models for Time Series Analysis: A Tutorial and Surveycited by 234
27 more
  1. 2018Task-Free Continual Learningcited by 219
  2. 2018Online Learning: A Comprehensive Surveycited by 203
  3. 2023FITS: Modeling Time Series with 10k Parameterscited by 185
  4. 2016Using Fast Weights to Attend to the Recent Pastcited by 161
  5. 2024Continual Learning for Large Language Models: A Surveycited by 144
  6. 2024Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Seriescited by 120
  7. 2015Online Optimization : Competing with Dynamic Comparatorscited by 99
  8. 2020Dynamic Regret of Convex and Smooth Functionscited by 90
  9. 2024VisionTS: Visual Masked Autoencoders Are Free-Lunch Zero-Shot Time Series Forecasterscited by 87
  10. 2023OneNet: Enhancing Time Series Forecasting Models under Concept Drift by Online Ensemblingcited by 71
  11. 2022Continual Learning with Foundation Models: An Empirical Study of Latent Replaycited by 61
  12. 2023Continual Learning: Applications and the Road Forwardcited by 59
  13. 2021Recent Advances of Continual Learning in Computer Vision: An Overviewcited by 55
  14. 2024Recent Advances of Foundation Language Models-based Continual Learning: A Surveycited by 46
  15. 2018Dynamical Isometry and a Mean Field Theory of RNNs: Gating Enables Signal Propagation in Recurrent Neural Networkscited by 44
  16. 2022Learning Fast and Slow for Online Time Series Forecastingcited by 43
  17. 2023Maintaining Plasticity in Deep Continual Learningcited by 42
  18. 2024An Analysis of Linear Time Series Forecasting Modelscited by 40
  19. 2024A Survey of Deep Learning and Foundation Models for Time Series Forecastingcited by 32
  20. 2025Battling the Non-stationarity in Time Series Forecasting via Test-time Adaptationcited by 32
  21. 2024Addressing Concept Shift in Online Time Series Forecasting: Detect-then-Adaptcited by 11
  22. 2023How Does It Function? Characterizing Long-term Trends in Production Serverless Workloadscited by 10
  23. 2024DAM: Towards A Foundation Model for Time Series Forecastingcited by 10
  24. 2021Evaluating Privacy-Preserving Machine Learning in Critical Infrastructures: A Case Study on Time-Series Classificationcited by 6
  25. 2023Approximate Bayesian Class-Conditional Models under Continuous Representation Shiftcited by 5
  26. 2024Serverless Cold Starts and Where to Find Themcited by 5
  27. 2023Chunking: Continual Learning is not just about Distribution Shiftcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 13.

14 papers.

Sort:

  1. 2026The Forecast After the Forecast: A Post-Processing Shift in Time Seriescited by 9
  2. 2026Learning the Context of Errors: Black-Box Online Adaptation of Time Series Foundation Modelscited by 5
  3. 2025Online time series prediction using feature adjustmentcited by 4
  4. 2026Post-Training in Time Series Foundation Models: A Unifying Frameworkcited by 3
  5. 2026TEFL: Prediction-Residual-Guided Rolling Forecasting for Multi-Horizon Time Seriescited by 2
  6. 2026Adapting Time Series Foundation Models through Data Mixturescited by 2
  7. 2026Learning from Yesterday's Error: An Efficient Online Learning Method for Traffic Demand Predictioncited by 1
  8. 2026What If We Let Forecasting Forget? A Sparse Bottleneck for Cross-Variable Dependenciescited by 1
  9. 2026PAC-Bayesian Reconstruction Guarantees for Time Series Variational Autoencoderscited by 1
  10. 2026Hedging Memory Horizons for Non-Stationary Prediction via Online Aggregationcited by 0
  11. 2026Beyond Average Error through Oracle-Informed Stress Tests for Time-Series Forecastingcited by 0
  12. 2026EPOC: Endpoint-Preserving Online Correction With Compressed Residual State for Multi-Horizon Time Series Forecastingcited by 0
2 more
  1. 2026Volatility-Clustering Adaptation for Financial Time Seriescited by 0
  2. 2026Latent Inference-Time Guidance of Time Series Foundation Modelscited by 0