pages/index.md

Repro - WIND: Weather Inverse Diffusion for Zero-Shot Atmospheric Modeling

Pages

Page
Claim 1: Single pre-trained WIND model replaces specialized baselines across diverse atmospheric tasks without task-specific fine-tuning
Claim 2: WIND solves probabilistic forecasting, spatial/temporal downscaling, sparse reconstruction, and conservation law enforcement as inverse problems via posterior sampling
Claim 3: Model generates physically consistent counterfactual extreme weather storylines under global warming scenarios
Conclusion
Reproduction poster