PEG-Tab: Sampling-Time Record Repair and Release Control for Tabular Synthesis

Pengfei Li; QinYi Liu; Mohammad Khalil

cs.LGSep 30, 2026

PEG-Tab: Sampling-Time Record Repair and Release Control for Tabular Synthesis

Authors: Pengfei Li, QinYi Liu, Mohammad Khalil

Abstract

Pretrained tabular generators can reproduce training records even when aggregate utility remains high. When retraining is unavailable or too costly, sampling and release are the remaining intervention points. We present PEG-Tab (Post-Training Energy Guidance for Tabular Synthesis), a post-training repair and release-control framework for frozen tabular generators. For each generated row, a generator-native operator creates two alternatives. A shared calibrated score compares the three candidates, favours lower-risk records, and applies a final release check. We instantiate this interface for GReaT, CTGAN, TVAE, and TabDDPM without updating their parameters. Across five datasets and four generator families, PEG-Tab reduces mean Near Copy from $0.078$ to $0.027$ and lowers aggregate Exact Copy to zero. Relative to a $3\times$ post hoc filter, it retains higher utility in 12 of 16 transfer settings and Pareto-dominates the filter in eight. Gains are concentrated in copy and proximity-related risks.

Figures & tables

Figure 1: PEG-Tab repairs and checks a generated record before release. Offline statistics define the record-level risk score. Each generator then uses its native sampling mechanism to produce alternatives. The same scoring and release rule is applied to the original row and the repaired candidates. The base generator is not retrained, and the score is not a formal privacy guarantee.

Term	Large when	Intended signal
$E_{\mathrm{disc}}$	Values or combinations are copied or rare	Direct reproduction and uniqueness
$E_{\mathrm{mix}}$	A candidate is close to a member or occupies a rare mixed-space region	Neighbourhood overlap and mixed-feature rarity
$E_{\mathrm{dens}}$	Training neighbours are closer than reference neighbours	Local concentration around training data

Table 1: Meaning of the record-level score components. Each term describes one candidate row and does not constitute a population-level privacy guarantee.

Generator	$\lambda$	Selection outcome
CTGAN	0.5	Meets both criteria
GReaT	1	Zero-copy fallback
TVAE	2	Minimum copy under utility constraint
TabDDPM	2	Best utility with zero copy

Table 2: Guidance strengths selected on the development dataset and then frozen.

Method	Utility $\uparrow$	Worst AUC $\downarrow$	TPR@1% $\downarrow$	Exact Copy $\downarrow$	Near Copy $\downarrow$
Vanilla	$0.561\pm 0.590$	$0.543\pm 0.040$	$0.011\pm 0.009$	$0.005\pm 0.019$	$0.078\pm 0.148$
Post hoc Filter	$\bm{0.608\pm 0.356}$	$0.543\pm 0.039$	$0.010\pm 0.009$	$\bm{0.000\pm 0.000}$	$0.051\pm 0.145$
PEG-Tab	$0.585\pm 0.433$	$0.536\pm 0.035$	$0.009\pm 0.008$	$\bm{0.000\pm 0.000}$	$\bm{0.027\pm 0.114}$
DP-Train	$0.023\pm 0.668$	$\bm{0.527\pm 0.024}$	$\bm{0.008\pm 0.008}$	$\bm{0.000\pm 0.001}$	$\bm{0.027\pm 0.068}$

Table 3: Aggregate results over five datasets, four generators, and three random seeds. Each entry reports the mean and standard deviation over 60 dataset, generator, and seed runs. The standard deviation reflects configuration and seed variability and is not a confidence interval. Utility is a descriptive cross-task average of classification AUC and regression

R^{2}

. Bold values denote the best. Worst AUC is the maximum AUC across Reconstruction, DCR, NN-Ratio, DPI, and DOMIAS.

Figure 2: PEG-Tab transfer effects relative to Vanilla after selecting $\lambda$ on South German Credit. Panel (a) reports the reduction in Near Copy, where positive values favour PEG-Tab. Panel (b) reports the change in task utility, where positive values favour PEG-Tab. South German Credit is excluded as it serves as the development dataset.

Figure 3: Absolute attack AUC deviation from chance. Points report the mean

|\mathrm{AUC}_{\mathrm{method}}-0.5|

, so values closer to zero indicate weaker attack discrimination. Bars show 95% configuration-bootstrap intervals. Worst AUC and the five score-based attacks use 20 dataset and generator cells. The held-out Shadow MIA row reports the selected-lambda transfer protocol on 12 cells with fixed shadow-release provenance.

Variant	Near $\downarrow$	Worst AUC $\downarrow$	Shadow $\downarrow$	Utility $\uparrow$
Vanilla	$0.0700$	$\bm{0.5410}$	$0.5059$	$0.6116$
Reject only	$0.0664$	$0.5432$	$0.5082$	$0.5566$
Rerank only	$0.0692$	$0.5425$	$\bm{0.5033}$	$0.5697$
Repair only	$0.0667$	$0.5418$	$0.5059$	$0.5873$
Repair + rank	$0.0609$	$0.5440$	$0.5065$	$\bm{0.6145}$
Full PEG-Tab	$\bm{0.0392}$	$0.5428$	$0.5070$	$0.6012$

Table 4: Pipeline ablation averaged over 20 dataset–generator cells. Lower values are better except for Utility. Worst is the maximum AUC over the five score-aligned attacks. Bold values denote the best result across all variants.

Generator	Utility wins	Near B/T/W	Pareto P/F/T
CTGAN	$4/4$	$3/0/1$	$3/0/1$
GReaT	$3/4$	$0/3/1$	$2/1/1$
TVAE	$4/4$	$1/1/2$	$2/0/2$
TabDDPM	$1/4$	$0/4/0$	$1/3/0$
All	$\bm{12/16}$	$\bm{4/8/4}$	$\bm{8/4/4}$

Table 5: Configuration-level comparison with Post hoc Filter using

3\times

oversampling on the 16 transfer cells. Utility wins count cells where PEG-Tab has higher Utility. Near B/T/W denotes better, tied, or worse Near Copy for PEG-Tab. Pareto P/F/T denotes PEG-Tab dominance, filter dominance, or a trade-off.

Method	$\varepsilon$	Utility $\uparrow$	Near $\downarrow$	Worst AUC $\downarrow$
PEG-Tab	n/a	$0.585$	$0.027$	$0.536$
MST	$1$	$0.307$	$\bm{0.023}$	$\bm{0.531}$
MST	$8$	$\bm{0.630}$	$0.026$	$0.551$
AIM	$1$	$0.213$	$0.025$	$0.546$
AIM	$8$	$0.462$	$0.032$	$0.551$

Table 6: Contextual comparison with formal DP reference mechanisms. The PEG-Tab row is taken from the three-seed benchmark in Table 3 . MST and AIM are evaluated on the same five datasets using the same utility and attack definitions. PEG-Tab has no formal

\varepsilon

. Bold values denote the best reported result in each column. Per-dataset results and privacy accounting details appear in the supplement.

Explore similar work

Self-Improving Tabular Language Models via Iterative Reward-Guided Post-Training
Apr 21, 2026Yunbo Long, Tejumade Afonja, Guangya Hao +2Post-Training Next-Token Distribution
Department of Engineering, University of Cambridge · CISPA Helmholtz Center for Information Security, Saarbrücken, Germany · The Alan Turing Institute, London
PATE-TabTransGAN: Differentially Private Synthetic Tabular Data Generation via Transformer-Based Student Discrimination
May 26, 2026M. Youssef, M. WoźniakSynthetic Tabular Data Privacy-Utility Trade-Offs
Wrocław University of Science and Technology, Wrocław, Poland
TabSCM: A practical Framework for Generating Realistic Tabular Data
Apr 24, 2026Sven Jacob, Bardh Prenkaj, Weijia Shao +1Synthetic Tabular Data Causal Discovery Methods
Federal Institute for Occupational Safety and Health (BAuA), Dresden, Germany · School of Computation, Information and Technology, Technical University of Munich, Munich, Germany · Munich Center for Machine Learning (MCML)

Apr 21, 2026, cs.LGJ/K move · Enter open · S save

Self-Improving Tabular Language Models via Iterative Reward-Guided Post-Training

Yunbo Long, Tejumade Afonja, Guangya Hao, Alexandra Brintrup+1

Department of Engineering, University of Cambridge · CISPA Helmholtz Center for Information Security, Saarbrücken, Germany · The Alan Turing Institute, London

Post-Training Next-Token Distribution Synthetic Data Large Language Model Generation Iterative

Open paper AskPDF arXiv

Tabular language models can generate synthetic tables by modeling rows as token sequences, but they are typically trained once with supervised fine-tuning and then used as static synthesizers. This is limiting because next-token likelihood does not directly optimize the distributional, utility, and indistinguishability properties used to evaluate synthetic data. We study iterative reward-guided post-training for tabular language models through a generate--score--align protocol, where a generator samples synthetic rows, a task-specified reward ranks them, and the model is updated relative to a fixed supervised reference. Within this protocol, we propose \textbf{TabGRAA} (\textbf{Tab}ular \textbf{G}roup-\textbf{R}elative \textbf{A}dvantage \textbf{A}lignment), a group-relative alignment method that compares high- and low-reward generated groups using group-averaged policy/reference log-ratios rather than one-to-one preference pairs. Across five mixed-type benchmarks, TabGRAA improves a GReaT backbone beyond additional supervised fine-tuning and achieves the strongest average trade-off among adapted DPO, KTO, and NPO baselines on fidelity and downstream utility, while maintaining empirical privacy diagnostics near the supervised baseline. Ablations show that the gains depend on meaningful reward ranking and stable group-level updates rather than extra training alone. Reward-substitution and scorer-separation studies further show that the post-training loop can use both classifier-based and classifier-free rewards, and that proper scorer separation is important for preserving the fidelity--utility--privacy trade-off. These results position TabGRAA as a self-improving post-training method for tabular language-model generators, complementary to strong static tabular synthesizers.

Abstract

Explore similar work

Self-Improving Tabular Language Models via Iterative Reward-Guided Post-Training

PATE-TabTransGAN: Differentially Private Synthetic Tabular Data Generation via Transformer-Based Student Discrimination

TabSCM: A practical Framework for Generating Realistic Tabular Data