Generating Benchmark Health Data Using a Tabular Diffusion Transformer

2026-08-14

AI Summary

This study addresses the challenges of heterogeneous modeling and low synthesis quality in cross-table data generation by proposing a two-stage framework based on statistical representation alignment. The method first converts heterogeneous tables into standardized statistical representations, then employs a diffusion Transformer combined with multivariate Gaussian sampling and inverse probability integral transforms to learn structural patterns. As a pioneering cross-table generation paradigm, this approach overcomes single-table limitations to enable unified modeling across unlimited heterogeneous tables. Experimental results demonstrate that the generated synthetic data achieves both high fidelity and diversity, effectively validating the method’s superiority in complex multi-source tabular generation tasks.

Abstract

Cross-Tabular Data Generation (CTDG) seeks to learn a generative model from multiple heterogeneous tables and produce new synthetic tabular datasets. However, existing synthetic tabular data generation methods are largely restricted to single-input-table scenarios and struggle to effectively handle multiple heterogeneous tables with diverse feature sets. To address this limitation, we propose a two-stage framework for cross-tabular data generation. In the first stage, each heterogeneous raw table is transformed into a standardized statistical table with the same set of columns across all tables. Each statistical table captures the marginal distributions of the original columns and the pairwise correlations among them. In the second stage, a diffusion transformer model is trained to capture structural patterns across these homogeneous statistical tables and to generate synthetic statistical tables. Synthetic raw tables are subsequently reconstructed from the generated statistical tables via multivariate Gaussian sampling followed by an inverse probability integral transform. This two-stage CTDG framework enables the learning of a unified generative model from multiple heterogeneous tables and supports the generation of an unlimited number of realistic synthetic heterogeneous tables. Experimental results demonstrate high fidelity in the learned statistical representations and a favorable fidelity-diversity trade-off in the generated synthetic data, validating the effectiveness of the proposed approach.

Problem

• Cross-Tabular Data Generation
• Heterogeneous Tables
• Synthetic Tabular Data
• Multi-table Generation

Innovation

• Cross-Tabular Data Generation
• Diffusion Transformer
• Heterogeneous Tables
• Statistical Representation
• Two-stage Framework