Reproduce “TabCUQ: Context-Uncertainty Quantification for 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
Yuebei Shi, Molei Liu, Zijian Guo
TabCUQ builds confidence intervals for conditional means and conditional average treatment effects from labeled contexts. It perturbs outcomes while keeping feature rows and model weights fixed, then uses estimate spread and the largest displacement as a bias cushion. In a TabPFN 2.6 quadratic simulation at x=3, coverage rose from 0.683 to 0.937, below nominal 0.95; foundation-model coverage is not guaranteed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.