A collection of processed datasets used in (https://arxiv.org/abs/2610.10398v1). Datasets larger than 1 TB are downsampled.