Reproduce “Efficient Provably Private Classification with a Tabular Foundation Model” 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
Talal Alrawajfeh, Cristiana Diaconu, Ossi Räisä, Sebastian Rodriguez Beltran, Yuan He, John Bronskill, Richard E. Turner, Antti Honkela
PrivTab enables private tabular classification by converting labelled context rows into a reusable noisy summary for prediction. Bounded attention with calibrated Gaussian noise creates the summary; later predictions use only this release and add no privacy loss under the stated guarantee. Across TabArena, PrivTab’s Elo was 1166.5 versus 1097.5 for DP-MLP, which outranked PrivTab under weak privacy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.