Research map / Probabilistic and scientific modeling / Causal discovery
TabStruct: Measuring Structural Fidelity of Tabular Data
Xiangjian Jiang, Nikola Simidjievski, Mateja Jamnik
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 diffusion generators (TabDDPM, TabSyn, TabDiff) instead of SMOTE | tabular generators · SCM datasets (6) | Global CI: 0.30 -> 0.69 | Improved |
| use SMOTE, which leads density estimation, instead of diffusion | SMOTE · SCM datasets (6) | Shape / Global CI: Shape 0.82 but Global CI 0.30 | Mixed |
| add structure-learning inductive bias (BN, GOGGLE) | tabular generators · SCM datasets (6) | Global CI: BN 0.35, GOGGLE 0.14 vs TabSyn 0.70 | No clear effect |
| use LLM autoregressive generation (GReaT) instead of SMOTE/TVAE | GReaT · Real-world datasets (23) | Global utility: 0.20 vs 0.41 (SMOTE) and 0.53 (TVAE) | Hurt |
| use local utility versus global utility as proxy for global CI | utility metric · SCM datasets (6) | Spearman correlation with global CI: 0.14 (local utility) vs 0.84 (global utility) | Hurt |
| aggregate utility over all features (global) rather than only the target (local) (ablation) | utility metric · SCM datasets (6) | Spearman correlation with structure: 0.78 (local utility vs local CI); 0.84 (global utility vs global CI) | Improved |
| use three untuned predictors (Tiny-default) instead of nine tuned predictors for global utility (ablation) | global utility · Real-world datasets (23) | evaluation time per 1000 samples: 0.64s vs 1.21s | Improved |
Closest papers by meaning
- Tabular Foundation Model for Generative Modelling (NeurIPS 2026 · similarity 0.87)
- RADAR: Benchmarking Language Models on Imperfect Tabular Data (NeurIPS 2025 · similarity 0.87)
- TabDiff: a Mixed-type Diffusion Model for Tabular Data Generation (ICLR 2025 · similarity 0.86)
- UniTabE: A Universal Pretraining Protocol for Tabular Foundation Model in Data Science (ICLR 2024 · similarity 0.86)
- Learning Causal Orderings for In-Context Tabular Prediction (NeurIPS 2026 · similarity 0.86)
- TabDPT: Scaling Tabular Foundation Models on Real Data (NeurIPS 2025 · similarity 0.85)
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.
