Sign inGet started
Dataset · Hugging Face

yamboo/RDB_PFN

RDB_PFN Datasets This repository contains the synthetic pre-training and benchmark datasets for RDB-PFN, as presented in the paper Relational In-Context Learning via Synthetic Pre-training with Structural Prior. Paper: https://arxiv.org/abs/2603.03805 GitHub Repository: https://github.com/MuLabPKU/RDBPFN Description RDB-PFN is the first relational foundation model trained purely via synthetic data. Because high-quality relational databases (RDBs) are often… See the full description on the dataset page: https://huggingface.co/datasets/yamboo/RDB_PFN.

Author / organizationyamboo
License (as declared)apache-2.0
Downloads287 (as reported by Hugging Face)
Likes0
Created12 March 2026
Last updated at source30 April 2026
tabularrelational-databasessynthetic-datain-context-learning
Embed this entry
<iframe src="https://silytics.si/embed/item/753bfc67-f584-4cf0-a69f-92feac8f0911" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Popularity, last 3 days

287 now (-4 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.

Connected in the knowledge graph

Explore this entry in the graph

Recent changes

  • 17 Oct 2026 Description: RDB_PFN Datasets This repository contains the synthetic pre-training and benchmark datasets for RDB-PFN, as presented in the paper Relational In-Context Learning via Synthetic Pre-training with Structural Prior. Paper: https://arxiv.org/abs → RDB_PFN Datasets This repository contains the synthetic pre-training and benchmark datasets for RDB-PFN, as presented in the paper Relational In-Context Learning via Synthetic Pre-training with Structural Prior. Paper: https://arxiv.org/abs

Changes detected by Silytics between refreshes of the source.

Related papers

More from yamboo

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.