analiza aiqbench
This research addresses the computational bottleneck of diffusion models in weather forecasting by introducing a distillation technique that compresses multi-step sampling into a single-step process. The method maintains high accuracy for extreme events like typhoons while significantly reducing neural function evaluations, offering a practical path toward scalable AI-driven climate prediction.
Diffusion Distillation for Efficient Weather Ensembles
arXiv:2608.27728v1 Announce Type: new Abstract: Diffusion models generate skillful weather ensembles but require costly iterative sampling. We introduce a supervised energy-distance distillation method that compresses a multi-step diffusion teacher into a single-step student by aligning student forecasts with teacher samples and ground-truth observations. Experiments on global forecasting and typhoon-track prediction show that our student outperforms existing distillation methods and preserves skill for extreme events. It matches or surpasses the teacher across key metrics using only one neural function evaluation per autoregressive step.
Abstract:Diffusion models generate skillful weather ensembles but require costly iterative sampling. We introduce a supervised energy-distance distillation method that compresses a multi-step diffusion teacher into a single-step student by aligning student forecasts with teacher samples and ground-truth observations. Experiments on global forecasting and typhoon-track prediction show that our student outperforms existing distillation methods and preserves skill for extreme events. It matches or surpasses the teacher across key metrics using only one neural function evaluation per autoregressive step.
Machine Learning (cs.LG); Applications (stat.AP)
arXiv:2608.27728 [cs.LG]
(or arXiv:2608.27728v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.27728 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 173 cuvinte
Subiecte conexe
Extreme Event Likelihoods with Guided Generative Models
NVIDIA · 13 iul. 2026
ORBIT-2 based Weather and Climate Downscaling and Downscaled Global Forecasts on AMD Instinct
AMD · 8 iun. 2026
Long-Range Distillation: Distilling 10,000 Years of Simulated Climate into Long Timestep AI Weather Models
arXiv · 15 oct. 2026
How can LLM RL Work Despite Information-Theoretic Inefficiency
LessWrong · 16 oct. 2026