Conference Open access

Generating High-Diversity Synthetic Tabular Data via Less-Constrained Prior

Sanghun Park Jaesung Lim Jong-June Jeon Seunghwan An

Sep 2026 · Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence · 0 citations · 45 references

TL;DR

This work empirically demonstrate that the model, which relies on a single prior distribution, significantly improves the quality of the synthetic data, especially in terms of diversity.

Abstract

Generating high-quality synthetic tabular data, regarding fidelity, diversity, and utility, is crucial for many practical purposes. Recent tabular data synthesis methods, including two-stage generative modeling approaches, have achieved this with nearly perfect fidelity. However, we observe that there remains room for improving diversity, and we show that this limitation arises from an additional constraint imposed on the latent support. Our main contribution is that we effectively eliminate this redundant constraint by directly deriving the objective function from the KL-divergence between the ground-truth density and the generative model used for synthetic sample generation. We empirically demonstrate that our model, which relies on a single prior distribution, significantly improves the quality of the synthetic data, especially in terms of diversity. Our implementation code is available at https://anonymous.4open.science/r/SPT-1EE2/.

DOI Semantic Scholar

Read PDF

Similar papers

Preprint Aug 2026

Limitations of Synthetic Data Generation in Specialized Data-Scarce Domains

This work focuses on domains where images differ substantially from common image datasets and additional data are expensive to obtain, and evaluates the downstream classifier performance improvements yielded by two schools of generative sparse data extension: distribution modeling and sample perturbation.

Edward Zhang, Marcel Hussing, Tanay Tandon et al. · 0 citations

Preprint Aug 2026

NAE: Normalizing AutoEncoder

This work proposes Normalizing Autoencoder (NAE), which employs a novel conditional loss that aligns the surrogate loss gradient with that of reconstruction loss, directly improving upon the current standard.

Muhammad Abdur Rafae, Niels Landwehr · 0 citations

On the Transfer of Output Diversity via Synthetic Data in Language Models

This paper finds that output-distribution tendencies can be partially transmitted through unrelated synthetic data, and suggests that diversity-related behavior is shaped, at least in part, by the latent model state rather than solely by decoding-time randomness.

Adisesh Venkatesh Sanklapur · 1 citation

#computer vision Jul 2026

ROMS-IMLE: A Minimalist Approach to Competitive Single-Step Generative Modelling

This work takes a minimalist approach to designing a competitive generative model that produces high quality samples at fast speed and achieves an FID of 2.56 on ImageNet 256 and simultaneously attains good precision and recall.

Chirag Vashist, Ke Li · arXiv.org · 0 citations

Preprint Aug 2026

Balanced Adaptive Prototype Selection for Scalable TabPFN Inference on Large-Scale Tabular Data

Experiments on the million-row HIGGS and SUSY datasets show that 512 prototypes retain strong predictive performance and reliable calibration, corresponding to an approximately 1,953-fold context compression.

M. Jadid, Melika Rezaye Garkani, A. Mousavi · 0 citations

#natural language process... Preprint Sep 2026

Debias-SparseGPT: Bias-Aware Pruning for Large Language Models

Debias-SparseGPT is introduced, a post-training pruning method incorporating representational debiasing using a second-order term defined over demographically contrasting inputs that consistently reduces pruning-induced bias compared to SparseGPT while preserving model perplexity and zero-shot accuracy.

Irina Proskurina, Guillaume Metzler, Antoine Gourru et al. · 0 citations