Research map / Image generation / Fast diffusion sampling
STaSy: Score-based Tabular data Synthesis
Jayoung Kim, Chaejeong Lee, Noseong Park
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 |
|---|---|---|---|
| Use score-based tabular synthesis with self-paced learning and fine-tuning | STaSy · 15-dataset average | quality (F1 & R²): 0.717 -> 0.733 | Improved |
| Use score-based tabular synthesis with self-paced learning and fine-tuning | STaSy · 15-dataset average | diversity (coverage): 0.637 -> 0.658 | Improved |
| Add self-paced learning and fine-tuning to Naïve-STaSy (ablation) | STaSy · 15-dataset TSTR average | F1: 0.777 -> 0.792 | Improved |
| Add self-paced learning and fine-tuning to Naïve-STaSy (ablation) | STaSy · 15-dataset average | runtime (second): 8.855 -> 10.663 | Hurt |
| Add fine-tuning (ablation) | STaSy · testing records, averaged over datasets | median log-probability: 129.293 -> 131.734 | Improved |
| Add self-paced learning (ablation) | STaSy · sampling diversity across datasets | coverage | Improved |
| Fine-tune with denoising score matching loss rather than Hutchinson estimation (ablation) | STaSy · Default | F1: 0.510 -> 0.523 | Improved |
| Fine-tune with denoising score matching loss rather than Hutchinson estimation (ablation) | STaSy · Contraceptive | F1: 0.438 -> 0.451 | Improved |
| Use probability flow instead of predictor-corrector reverse-SDE solvers (ablation) | Naïve-STaSy · Magic | F1: AS: 0.752 -> 0.777; RD: 0.779 -> 0.775; EM: 0.779 -> 0.771 | Mixed |
| Add self-paced learning to STaSy oversampling (ablation) | STaSy w/o fine-tuning · HTRU, Magic, and Robot | Weighted-F1: 0.8767 -> 0.8803; 0.7807 -> 0.7960; 0.9031 -> 0.9270 | Improved |
Later papers that compared with this method
3 later accepted papers report STaSy as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Understanding and Mitigating Memorization in Diffusion Models for Tabular Data (ICML 2025 · better on 1 of 1 shared result)
- Mixed-Type Tabular Data Synthesis with Score-based Diffusion in Latent Space (ICLR 2024 · better on 2 of 2 shared results)
- CoDi: Co-evolving Contrastive Diffusion Models for Mixed-type Tabular Synthesis (ICML 2023 · better on 5 of 5 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TabSDS: a Lightweight, Fully Non-Parametric, and Model Free Approach for Generating Synthetic Tabular Data (ICML 2025 · similarity 0.89)
- CTSyn: A Foundation Model for Cross Tabular Data Generation (ICLR 2025 · similarity 0.88)
- Continuous Diffusion for Mixed-Type Tabular Data (ICLR 2025 · similarity 0.88)
- Mixed-Type Tabular Data Synthesis with Score-based Diffusion in Latent Space (ICLR 2024 · similarity 0.87)
- STRABLE: Benchmarking Tabular Machine Learning with Strings (NeurIPS 2026 · similarity 0.87)
- Tabular Foundation Model for Generative Modelling (NeurIPS 2026 · 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.
