A rigorous evaluation shows that simple Probabilistic Circuits often outperform complex diffusion-based models for tabular data generation at a fraction of the cost.
March 25, 2026
Original Paper
A Sobering Look at Tabular Data Generation via Probabilistic Circuits
arXiv · 2603.23016
The Takeaway
It reveals that the current hype surrounding tabular diffusion is largely driven by inadequate metrics. For practitioners, this suggests that classical mixture-model approaches are not only faster but more reliable for generating heterogeneous tabular data, challenging the current SOTA narrative.
From the abstract
Tabular data is more challenging to generate than text and images, due to its heterogeneous features and much lower sample sizes. On this task, diffusion-based models are the current state-of-the-art (SotA) model class, achieving almost perfect performance on commonly used benchmarks. In this paper, we question the perception of progress for tabular data generation. First, we highlight the limitations of current protocols to evaluate the fidelity of generated data, and advocate for alternative o