# 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-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-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](#/claim-3-model-generates-physically-consistent-counterfactual-extreme-weather-storylines-under-global-warming-scenarios) |
| [Conclusion](#/conclusion) |
| [Reproduction poster](#/reproduction-poster) |