Synthetic-to-Real Data Ratios in Pretraining Enhance Robustness of Tabular Foundation Models

This report synthesises findings from 8 peer-reviewed papers addressing the following research question: How does the ratio of synthetic-to-real data in pretraining affect the robustness of tabular foundation models (TFMs) on adversarial perturbations, as measured by accuracy degradation on TabBench OOD. 7 claims were extracted from source literature; 7 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.1/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: How does the ratio of synthetic-to-real data in pretraining affect the robustness of tabular foundation models (TFMs) on adversarial perturbations, as measured by accuracy degradation on TabBench OOD sets with added noise?

Autonomous literature synthesis. Automated review score: 8.1/10. Full text and citation available at Assignee Research.