Parameter-Efficient Distributionally Robust Adaptation of Tabular Foundation Models under Subpopulation Shift

Excerpt

arXiv:2610.01143v1 Announce Type: new Abstract: Despite strong mean accuracy, tabular foundation models (TFMs) can perform poorly on underrepresented groups under subpopulation shift, where group proportions change between training and deployment. We propose DR-TFM, a parameter-efficient distributionally robust adaptation framework that requires no true group annotations. DR-TFM adjusts attention to labeled context examples by fine-tuning an existing query scaling network or adding and training