Reproduce “TabDAG: Structured Evolutionary Search for Tabular Feature Engineering” 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

Anastasia Volkova, Andrey Kuznetsov, Ivan Oseledets, Valentin Khrulkov

TabDAG represents tabular feature programs as typed dependency graphs and uses parent-specific schemas to guide language-model edits, with checks before candidate evaluation. Its features scored above raw inputs in 26 of 30 dataset–estimator cells, with fewer failed evaluations than Python and whole-graph JSON rewrites. Evidence covers ten datasets and one primary search per method and cell.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.