Research map / Image generation / Fast diffusion sampling
AugMask: Training Diffusion Models on Incomplete Tabular Data via Stochastic Augmentation and Masking
Jungkyu Kim, Taeyoung Park, Kibok Lee
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 conditional stochastic augmentation (AugFull) to missing-unaware generators | TabSyn · MCAR 90 scenarios | Summary rank: 13.7 -> 6.3 | Improved |
| add AugFull to TVAE | TVAE · MCAR 90 scenarios | Summary rank: 17.7 -> 11.2 | Improved |
| add AugFull to ARF | ARF · MCAR 90 scenarios | Summary rank: 18.5 -> 12.4 | Improved |
| add observed-only loss masking on top of augmentation (AugMask vs AugFull) (ablation) | TabDiff · MCAR 90 scenarios | Summary rank: 5.2 -> 3.2 | Improved |
| add observed-only loss masking on top of augmentation (ablation) | CDTD · MCAR 90 scenarios | Summary rank: 5.8 -> 3.8 | Improved |
| add observed-only loss masking on top of augmentation (ablation) | TabDDPM · MCAR 90 scenarios | Summary rank: 7.5 -> 7.0 | Improved |
| use stochastic instead of deterministic conditional augmentation (ablation) | AugMask · Adult and others | Quality | Improved |
| vary augmenter context bias and stochasticity (misspecification) (ablation) | AugMask vs AugFull · 5x5 sweep of eta and lambda | Delta Quality | Improved |
| cycle among cached augmentations instead of one fixed draw (ablation) | LGBM-S augmentation · Adult beta-Recall | beta-Recall: 0.3305 -> 0.1553 | Hurt |
| cycle among cached augmentations (ablation) | LGBM-S augmentation · Adult MLE utility | MLE: 0.8877 -> 0.8924 | No clear effect |
| mix a fixed anchor (80%) with cycling (ablation) | LGBM-S augmentation · Adult beta-Recall | beta-Recall: 0.1679 -> 0.2724 | Mixed |
| use AugMask for CDTD under heavy missingness | CDTD · Adult pmiss=0.7 | UMAP IoU: 0.1574 -> 0.9285 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DiffPuter — DiffPuter: Empowering Diffusion Models for Missing Data Imputation (ICLR 2025) · better on 1 of 1 shared result
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
- TabDiff: a Mixed-type Diffusion Model for Tabular Data Generation (ICLR 2025 · similarity 0.88)
- Learning What Not to Impute: An Uncertainty-Aware Diffusion Framework for Meaningful Missingness (NeurIPS 2026 · similarity 0.87)
- ReMasker: Imputing Tabular Data with Masked Autoencoding (ICLR 2024 · similarity 0.87)
- To Predict or Not to Predict? Proportionally Masked Autoencoders for Tabular Data Imputation (AAAI 2025 · similarity 0.87)
- Rethinking the Diffusion Models for Missing Data Imputation: A Gradient Flow Perspective (NeurIPS 2024 · similarity 0.87)
- Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking (NeurIPS 2025 · 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.
