GenCFD: Representative samples from each dataset

This record accompanies the paper Generative AI for Efficient Statistical Computation of Fluids, and the associated code release (Zenodo DOI: 10.5281/zenodo.20744140). In the Main Text we present results on five challenging three-dimensional flow datasets. Because of the severe per-record size limits of Zenodo, this record provides 10 representative trajectories from each dataset rather than the full data. The complete datasets (all trajectories, all macroscopic conditions) are available under a CC BY 4.0 license in a Google Cloud Storage bucket: Naming convention in this record - `<Name>.nc` — 10 trajectories (`member = 10`) from the primary dataset.- `<Name>_single_macro_test.nc` — a single macro sample from the micro–macrodataset, with 10 microscopic realizations. All field variables are float32Physical coordinate arrays (x, y, z, time) are stored in every file Datasets: 1. Taylor–Green vortex — TaylorGreen*.nc2. Cylindrical shear flow — CylindricalShearFlow*.nc3. Cloud–shock interaction — CloudShock*.nc4. Nozzle flow — NozzleFlow*.nc5. Dry convective planetary boundary layer — DryConvectivePlanetaryBoundaryLayer*.nc