Impact of CausalMixFT Synthetic Data Ratios on Downstream TFM Accuracy Across Model Sizes on HouseElec

Synthetic tabular data generation addresses data scarcity and privacy constraints in a variety of domains. Tabular Prior-Data Fitted Network (TabPFN), a recent foundation model for tabular data, has been shown capable of generating high-quality synthetic tabular data. However, TabPFN is autoregressive: features are generated sequentially by conditioning on the previous ones, depending on the order in which they appear in the input data. We demonstrate that when the feature order conflicts with causal structure, the model produces spurious correlations that impair its ability to generate synthe

Research goal: How does scaling the synthetic data ratio (1:1, 2:1, 3:1) generated by CausalMixFT impact the downstream accuracy of fine-tuned TFMs on the HouseElec benchmark, and does this hold across different model sizes (e.g., TABRIEF vs. TabPFN)?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 9.0/10.