Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 56 min ago
Peer-reviewedCC BYMachine Learning Earth

Swift: an autoregressive consistency model for efficient weather forecasting

  • Jonathan G. Stock1
  • Troy Arcomano2,1
  • Veerabhadra Rao Kotamarthi1
  1. 1Argonne National Laboratory
  2. 2Allen Institute for AI
Open at sourceDOICite

Abstract

Diffusion models offer a physically grounded framework for probabilistic weather forecasting, but their typical reliance on slow, iterative solvers during inference makes them impractical for subseasonal-to-seasonal (S2S) applications where long lead-times and domain-driven calibration are essential. To address this, we introduce Swift, a single-step consistency model that, for the first time, enables autoregressive finetuning of a probability flow model with a continuous ranked probability score (CRPS) objective. This eliminates the need for multi-model ensembling or parameter perturbations. Results show that Swift produces skillful 6-hourly forecasts that remain stable for up to 75 days, running 39× faster than state-of-the-art diffusion baselines while achieving forecast skill competitive with the numerical-based, operational IFS ENS. This marks a step toward efficient and reliable ensemble forecasting from medium-range to seasonal-scales.

Related

  1. PhysMetrics.Weather: An Evaluation Framework for Physical Consistency in ML Weather Models
    Emma Kasteleyn, Timo Maier, Axel Lauer +3
    arXiv (Cornell University)PreprintLicence not statedFull text6 citedAtmospheric science & meteorologyOpenAlex
  2. FastNet: Improving the Physical Consistency of Machine Learning Weather Prediction Models through Loss Function Design
    Tom Dunstan, Oliver Strickson, Thusal Bennett +34
    Artificial Intelligence for the Earth SystemsPeer-reviewedLicence not statedFull text10 citedAtmospheric science & meteorologyOpenAlex
  3. SDECast: Probabilistic Weather Forecasting in Continuous Time with Neural SDEs
    Maria Marchenko, Martin Andrae, Fredrik Lindsten +1
    arXivPreprintarXiv licenceFull textAtmospheric science & meteorologyStatisticsarXiv
  4. Butterfly Effect Confirmed in Global AI Weather Models: Evidence from Tropical Cyclone Forecasting
    Jeremy Cheuk-Hin Leung, Daosheng Xu, Weiye Yu +9
    arXivPreprintCC BY-NC-SA 4.0Full textAtmospheric science & meteorologyStatistical & nonlinear physicsarXiv
  5. Weather forecasting in a changing climate: the rise of AI and Machine learning?
    Florence Rabier, Andrew Brown, Matthew Chantry +1
    Journal of the European Meteorological Society.Peer-reviewedCC BY7 citedAtmospheric science & meteorologyOpenAlex

Cite

BibTeXRIS
BibTeX
@article{stock2026swift,
  title = {{Swift: an autoregressive consistency model for efficient weather forecasting}},
  author = {Jonathan G. Stock and Troy Arcomano and Veerabhadra Rao Kotamarthi},
  journal = {Machine Learning Earth},
  year = {2026},
  month = jun,
  doi = {10.1088/3049-4753/ae802a},
  url = {https://doi.org/10.1088/3049-4753/ae802a},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Swift: an autoregressive consistency model for efficient weather forecasting
AU  - Jonathan G. Stock
AU  - Troy Arcomano
AU  - Veerabhadra Rao Kotamarthi
PY  - 2026
DA  - 2026/06/22
JO  - Machine Learning Earth
DO  - 10.1088/3049-4753/ae802a
UR  - https://doi.org/10.1088/3049-4753/ae802a
AB  - Diffusion models offer a physically grounded framework for probabilistic weather forecasting, but their typical reliance on slow, iterative solvers during inference makes them impractical for subseasonal-to-seasonal (S2S) applications where long lead-times and domain-driven calibration are essential. To address this, we introduce Swift, a single-step consistency model that, for the first time, enables autoregressive finetuning of a probability flow model with a continuous ranked probability score (CRPS) objective. This eliminates the need for multi-model ensembling or parameter perturbations. Results show that Swift produces skillful 6-hourly forecasts that remain stable for up to 75 days, running 39× faster than state-of-the-art diffusion baselines while achieving forecast skill competitive with the numerical-based, operational IFS ENS. This marks a step toward efficient and reliable ensemble forecasting from medium-range to seasonal-scales.
SN  - 3049-4753
ER  -