AI RESEARCH

Benchmarking Tabular Foundation Models for Conditional Density Estimation in Regression

arXiv CS.LG

ArXi:2603.26611v1 Announce Type: new Conditional density estimation (CDE) - recovering the full conditional distribution of a response given tabular covariates - is essential in settings with heteroscedasticity, multimodality, or asymmetric uncertainty. Recent tabular foundation models, such as TabPFN and TabICL, naturally produce predictive distributions, but their effectiveness as general-purpose CDE methods has not been systematically evaluated, unlike their performance for point prediction, which is well studied.