Generalization of Adversarially Pretrained Tabular Foundation Models on Real-World Benchmarks

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: To what extent do tabular foundation models pretrained on synthetic data with adversarial noise generalize to real-world tabular datasets, as evaluated by accuracy comparisons on benchmarks like TabMNAR and TabCI?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 7.5/10.