Empirical Evaluation of Out-Of-Distribution Performance of Tabular Foundation Models
Malena Loza, David Chushig-Muzo, Eva Milara, Luis Bote-Curiel, Luis Estrada-Petrocelli, Felipe Grijalva
Why It Matters
What makes this one worth your time
Understanding the OOD performance of TFMs is crucial for their reliable deployment in real-world applications where distribution shifts are common.
The study evaluates the robustness of tabular foundation models under distribution shifts, revealing performance degradation and scalability challenges.
Summary
The paper conducts an empirical evaluation of the Out-Of-Distribution (OOD) performance of nine Tabular Foundation Models (TFMs) across three real-world datasets, highlighting their systematic degradation under distribution shifts and identifying a scalability gap in resource demands.
Key contributions
- Empirical evaluation of nine TFMs under distribution shifts.
- Identification of performance degradation patterns across different shift types.
- Highlighting the scalability gap in resource requirements for high-performing TFMs.
Notable insights
- All evaluated TFMs show systematic performance degradation under distribution shifts.
- High-performing TFMs require significant computational resources, posing scalability challenges.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2607.26000v1 Announce Type: cross Abstract: Tabular Foundation Models (TFMs) have emerged as novel approaches for tabular predictive tasks, demonstrating competitive predictive performance to ensemble tree-based models. Most TFMs are trained and evaluated on independent and identically distributed data, but this assumption changes in real-world scenarios due to distribution shifts, which compromise the robustness of models. Limited research has been conducted of TFMs under distribution shifts. We present an empirical evaluation of Out-Of-Distribution (OOD) performance of nine TFMs, spanning diverse pre-training strategies and architectures: TabPFNv2, TabPFNv2.5, TabPFNv2.6, TabPFNv3, TabICL, TabICLv2, Mitra, LimiX and TabFM. Three real-world datasets from the TableShift study were considered (HELOC, Voting, Childhood Lead), covering label, socioeconomic, and geographic shift types. Our results show that all evaluated TFMs degrade systematically under distribution shift regardless of pre-training strategy, with shift gaps ranging from 0.003 to 0.060 depending on shift type. The relationship between in-distribution and OOD predictive performance documented for classical tabular models extends into TFMs. We also identified a scalability gap, as high-performing models demand significant memory and computational resources beyond what standard deployment infrastructure can support. This study extends existing benchmarks for OOD in tabular data, providing evidence to support their adoption in high-stakes domains characterized by structural distribution shifts.