Reproduce “CoRTeC: A Differentially Private Release for Synthetic Tabular Data, Decodable by Any Generator” 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

Calister Nnona

CoRTeC lets institutions publish noisy counts and generate synthetic records from them without exposing private rows to a language model. Its fixed release satisfies pure ε-differential privacy; decoding and repeated draws use no additional privacy budget. On Adult, selected maximum-entropy synthetic data trained classifiers within 0.02 test AUC of same-size real samples across three splits, at 300 and 1,000 rows.

Summary by alphaXiv.

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