Reproduce “Why Tabular Foundation Models Excel on Relational Data They Have Never Seen” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Linjie Xu, Yiqun Hu, Jixiang He, David Wipf
RDBLearn pairs a parameter-free relational encoder with a tabular foundation model to predict from relational databases. The encoder summarizes related rows as column-aligned features, which the model uses with labeled examples at inference time. On eight SALT tasks, it reports average MRR of 0.88 versus 0.82 for KumoRFM-2, an uncontrolled benchmark comparison, not evidence of superiority over all supervised models.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.