Time Series Foundation Models
Investigates pretraining strategies for building foundation models that generalize across multivariate time series forecasting tasks.
5 papers
Where this stands
The written synthesis of this thread is for subscribers. Subscribe.
Results across this thread
6 reported results from the papers in this thread.
The table is for subscribers. Subscribe to see every reported number side by side.
How this thread developed
Oct 2026 · Tsinghua University, ByteDance
Pretraining on synthetic primitive patterns enables robust zero-shot multivariate time series forecasting
Introduces Timer-M1, a multivariate time series foundation model pretrained on synthetic primitive patterns for zero-shot forecasting.
1 further paper
Nov 2026 · Amazon Web Services, University of Freiburg
AI forecasting model can rival specialist tools without dataset-specific training
Introduces Chronos-2, a pretrained forecaster that models related measurements jointly without dataset-specific training, advancing zero-shot time series foundation models.
1 further paper
Nov 2026 · Indian Institute of Technology Patna, Sopra Banking Software India
AI forecasting model adapts to sensor changes, guarantees uncertainty at new locations
Proposes a topology-invariant spatiotemporal forecaster with conformal uncertainty calibration, advancing foundation models for time series.
released code
3 of 5 papers shown