Dataset for: Learning Transient Convective Heat Transfer with Geometry Aware World Models

This dataset accompanies the paper "Learning Transient Convective Heat Transfer with Geometry Aware World Models". https://arxiv.org/abs/2601.22086 The dataset contains high-resolution CFD simulations of transient convective heat transfer for a parametrized geometry. The data is provided both as raw simulation outputs (HDF5 format) and as processed tensor datasets suitable for machine learning workflows. Processed datasets include training and testing splits used in the experiments described in the paper. Scripts used to generate the processed tensors are included for reproducibility.