Research map / Large language models / AI and society
Reimagining Synthetic Tabular Data Generation through Data-Centric AI: A Comprehensive Benchmark
Lasse Hansen, Nabeel Seedat, Mihaela van der Schaar, Andrija Petrovic
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Cleanlab postprocessing (remove hard samples) to bayesian_network synthetic data (ablation) | bayesian_network · classification | AUROC % change: 0.31 | No clear effect |
| add Cleanlab pre- and postprocessing to nflow (ablation) | nflow · classification | AUROC % change: 1.64 | Improved |
| add easy_hard preprocessing alone to ctgan (ablation) | ctgan · classification | AUROC % change: -0.78 | Hurt |
| add easy_hard preprocessing alone to tvae (ablation) | tvae · classification | AUROC % change: -0.25 | Hurt |
| add easy_hard preprocessing to bayesian_network (ablation) | bayesian_network · model selection | Spearman % change: 92.18 | Improved |
| add data-centric processing to tvae (ablation) | tvae · model selection | Spearman % change: -3.58 | Hurt |
| add pre- and postprocessing to ctgan (ablation) | ctgan · model selection | Spearman % change: 14.73 | Improved |
| add data-centric processing across generative models (ablation) | all models · statistical fidelity | inverse KL | Hurt |
| add data-centric processing, 15 evaluations (ablation) | all models · classification | AUROC: 13 out of 15 | Improved |
| add data-centric processing, 15 evaluations (ablation) | all models · feature selection | Spearman: 12 out of 15 | Improved |
| increase label noise above 8% on TabDDPM Covid dataset | data-centric processing · Covid mortality | performance gain | Hurt |
| choose bayesian_network over other generators by statistical fidelity | bayesian_network · 11 datasets | inverse KL vs AUROC: 0.998 vs 0.622 | Mixed |
Closest papers by meaning
- TabSDS: a Lightweight, Fully Non-Parametric, and Model Free Approach for Generating Synthetic Tabular Data (ICML 2025 · similarity 0.89)
- A Data-Centric Perspective on Evaluating Machine Learning Models for Tabular Data (NeurIPS 2024 · similarity 0.88)
- How Realistic Is Your Synthetic Data? Constraining Deep Generative Models for Tabular Data (ICLR 2024 · similarity 0.88)
- Synthetic Tabular Data Generation for Imbalanced Classification: The Surprising Effectiveness of an Overlap Class (AAAI 2025 · similarity 0.87)
- Generative Forests (NeurIPS 2024 · similarity 0.87)
- EPIC: Effective Prompting for Imbalanced-Class Data Synthesis in Tabular Data Classification via Large Language Models (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
