Reproduce “Evaluation of Active Feature Acquisition Policies with Tabular Foundation Models” 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

Yuta Kobayashi, Divyam Madaan, Shalmali Joshi

The authors develop a method to estimate active feature acquisition policy values from historical data. Using prior-data fitted networks, it averages entropy across posterior predictive draws to separate outcome variability from uncertainty about the predictor. In synthetic experiments, the method reduces value-estimation bias and improves policy ranking, especially under high missingness, but does not improve ranking in real-world Sepsis data.

Summary by alphaXiv.

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