A supervised energy-distance distillation method is introduced that compresses a multi-step diffusion teacher into a single-step student by aligning student forecasts with teacher samples and ground-truth observations and preserves skill for extreme events.
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.
GEM-3, a probabilistic global weather model that addresses this trade-off through explicit multi-timestep inference, is presented, a practical forecasting system that couples near-SOTA medium-range probabilistic skill, stable extended-range rollouts, efficient training and inference, and decision-relevant diagnostics.
Sam Levang, Fran Bartolić, Ty A. Dickinson et al.· 0 citations
Results indicate that cross-attention conditioning offers advantages over simple concatenation for probabilistic precipitation downscaling, and that pre-trained foundation model representations may offer benefits in data-limited settings.
Victor Nascimento Ribeiro, Jorge Guevara, J. Moraga et al.· 0 citations
The diffusion ensemble, a form of generative artificial intelligence, is used to make ensemble forecasts consisting of 1000 members from single-member deterministic dynamical West-WRF forecasts over the North Pacific, which demonstrated utility in representing real outcomes of extreme destructive events with realistic...
Timothy B. Higgins, William E. Chapman, A. Subramanian et al.· Artificial Intelligence for...· 0 citations
Scenario planning for rare, high-impact events often requires massive ensembles to stochastically sample relevant trajectories. Although autoregressive weather emulators can efficiently generate such ensembles, isolating trajectories of interest requires sifting through petabytes of data, a challenge that grows exponen...
Jerry Lin, M. Chien, Mansi Sakarvadia et al.· 1 citation
Stochastically perturbed weights (SPW) is at present a tuning procedure rather than a plug-and-play recipe, and its main failure mode is a coherent whole-field offset that overdisperses the domain mean, and restricting the noise to coarse scales or perturbing the initial conditions each repair part of it.
Simon Adamov, O. Fuhrer, R. Knutti et al.· 0 citations
Experimental results show that LoRA-EnVar significantly improves assimilation accuracy over models assuming static background error distribution and achieves comparable or better performance than full finetuning while reducing the number of trainable parameters by three orders of magnitude.
Yi Xiao, Hang Fan, Kun Chen et al.· Neural Information Processin...· 3 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 14, 2026
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
MIT News · Artificial Intelligence· news.mit.eduOct 13, 2026
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
