CTU relational datasets (redelex)

71 databases from the CTU Prague Relational Learning Repository in RelBench format, ported via redelex. Labels are redelex's task definitions served as-is (kind: external); *-original tasks use a fixed random 80/10/10 entity split, *-temporal tasks split by time. The dataset viewer lists every database and task.

<dataset>/
  manifest.yaml                 # tables, primary keys, foreign-key graph, val/test timestamps
  schema.svg                    # ER diagram
  db/*.parquet                  # relational tables
  tasks/<task>/manifest.yaml    # task spec
  tasks/<task>/{train,val,test}.parquet

import relbench
ds = relbench.load_dataset("stanford-star/redelex/ctu-financial")
task = relbench.load_task("stanford-star/redelex/ctu-financial", "<task>")

License

Each database keeps the license of its upstream source, listed in the license and source_url columns of STATS/databases.parquet. Our contributions (the conversion to this format, schemas, task definitions, splits and statistics) are CDLA-Permissive-2.0.

Citation

@misc{peleska2025redelex,
  title={REDELEX: A Framework for Relational Deep Learning Exploration},
  author={Jakub Peleška and Gustav Šír},
  year={2025},
  eprint={2506.22199},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2506.22199},
}

@inproceedings{relbenchv2,
  title={RelBench v2: A Large-Scale Benchmark and Repository for Relational Data},
  author={Gu, Justin and Ranjan, Rishabh and Kanatsoulis, Charilaos and Tang, Haiming and Jurkovic, Martin and Hudovernik, Valter and Znidar, Mark and Chaturvedi, Pranshu and Shroff, Parth and Li, Fengyu and Leskovec, Jure},
  booktitle={3rd Workshop on Navigating and Addressing Data Problems for Foundation Models (DATA-FM) at ICLR 2026},
  year={2026},
  note={arXiv:2602.12606 [cs.LG]},
  url={https://arxiv.org/abs/2602.12606}
}