Divergent Failure Modes of Synthetic-Pretrained Tabular Foundation Models Under Structured Adversarial Noise on TabBench

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 tabular foundation models pretrained with high synthetic data ratios exhibit different failure modes compared to real-data-trained models on TabBench tasks under structured adversarial noise?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 8.7/10.