Calibration Performance of Synthetic-Pretrained Multimodal Tabular Foundation Models on Perturbed OOD TabBench Subsets

The development of tabular foundation models (TFMs) has accelerated in recent years, showing strong potential to outperform traditional ML methods for structured data. A key finding is that TFMs can be pretrained entirely on synthetic datasets, opening opportunities to design data generators that encourage desirable model properties. Prior work has mainly focused on crafting high-quality priors over generators to improve overall pretraining performance. Our insight is that parameterizing the generator distribution enables an adversarial robustness perspective: during training, we can adapt the

Research goal: Do multimodal tabular foundation models pretrained on synthetic data retain calibration performance better than real-data-only models when evaluated on perturbed OOD subsets of TabBench?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 8.5/10.