TabFORGE: A Tabular Foundation Model for Generative Modelling
TabFORGE is a novel tabular foundation model designed for generating high-quality synthetic tabular data by focusing on causal structures. It uses a two-stage design involving a causality-aware encoder and a score-based diffusion transformer to learn robust, generalizable representations from diverse datasets. This approach results in synthetic data with strong structural fidelity and superior causal preservation compared to existing methods.