Pith. sign in

REVIEW 11 cited by

REaLTabFormer: Generating Realistic Relational and Tabular Data using Transformers

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2302.02041 v1 pith:MJ4RGLWN submitted 2023-02-04 cs.LG

Aivin V. Solatorio , Olivier Dupriez This is my paper
classification cs.LG
keywords relationaldatatabulardatasetsmodelrealtabformerparenttable
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

Tabular data is a common form of organizing data. Multiple models are available to generate synthetic tabular datasets where observations are independent, but few have the ability to produce relational datasets. Modeling relational data is challenging as it requires modeling both a "parent" table and its relationships across tables. We introduce REaLTabFormer (Realistic Relational and Tabular Transformer), a tabular and relational synthetic data generation model. It first creates a parent table using an autoregressive GPT-2 model, then generates the relational dataset conditioned on the parent table using a sequence-to-sequence (Seq2Seq) model. We implement target masking to prevent data copying and propose the $Q_{\delta}$ statistic and statistical bootstrapping to detect overfitting. Experiments using real-world datasets show that REaLTabFormer captures the relational structure better than a baseline model. REaLTabFormer also achieves state-of-the-art results on prediction tasks, "out-of-the-box", for large non-relational datasets without needing fine-tuning.

Discussion (0). Sign in to comment.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. TabQueryBench: A Query-Centric Benchmark for Synthetic Tabular Data

    cs.DB 2026-07 accept novelty 7.0 of 10

    Across 49 datasets and 11 generators, distance-based fidelity overstates synthetic tabular quality: best query-centric score is only 0.75, with systematic failures on high-cardinality support, local conditionals, and ...

  2. LAB-Tab: LLM-Augmented Bayesian Network Adaptation for Few-Shot Tabular Generation

    cs.LG 2026-08 conditional novelty 6.0 of 10

    A source-trained Bayesian network, augmented with LLM-proposed edges and calibrated by PPO, generates synthetic tabular data that outperforms six baselines in six ACS distribution-shift scenarios.

  3. TopoFE: topology-aware LLM-guided Automated Feature Engineering

    cs.AI 2026-07 conditional novelty 6.0 of 10

    TopoFE, a multi-island LLM-guided evolutionary feature engineering method with learned topology-aware transfer, outperforms AutoFE baselines on most of 29 tabular datasets.

  4. Do Generative Models Keep Time? A Time-Aware Evaluation of Synthetic Sequential Tabular Data

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Static-distribution fidelity is a poor proxy for temporal fidelity in synthetic sequential tabular data; measuring timestamp, trajectory, cross-sectional, and relational structure over time changes model rankings.

  5. ASPEN: An Additional Sampling Penalty Method for Finite-Sum Optimization Problems with Nonlinear Equality Constraints

    math.OC 2025-08 unverdicted novelty 6.0 of 10

    An abstract announces a new sampling penalty method with an almost sure convergence proof, but the manuscript body is a different paper, leaving the central claim unverifiable.

  6. Taming Data Challenges in ML-based Security Tasks Using Generative AI

    cs.CR 2025-07 conditional novelty 6.0 of 10

    Generative AI data augmentation, especially Nimai's sample-conditioned synthesis, improves several security classifiers and speeds drift recovery, but fails on tasks with noisy or overlapping labels.

  7. IMAGIC-500: IMputation benchmark on A Generative Imaginary Country (500k samples)

    cs.LG 2025-06 conditional novelty 6.0 of 10

    IMAGIC-500 is a large synthetic socioeconomic dataset with nested household structure, used to benchmark 14 imputation methods under MCAR, MAR, and MNAR at 10-50% missingness.

  8. A Survey on Evaluating Quality and Trustworthiness in LLM-Generated Data

    cs.AI 2026-01 conditional novelty 5.0 of 10

    A metric-oriented survey that classifies intrinsic quality and trustworthiness metrics for LLM-generated data across six modalities and documents systematic evaluation gaps in the current literature.

  9. Privacy-Preserving Tabular Synthetic Data Generation Using TabularARGN

    cs.LG 2025-08 unverdicted novelty 5.0 of 10

    TabularARGN is a discretization-based auto-regressive network claimed to generate high-fidelity, privacy-robust synthetic tabular data, competitive with diffusion and GAN baselines.

  10. The Prompt is Mightier than the Example

    cs.LG 2025-05 conditional novelty 5.0 of 10

    Injecting domain knowledge into prompts can substitute for many in-context examples in LLM-based synthetic tabular data generation, cutting required example counts by 40-90%.

  11. Towards High Supervised Learning Utility Training Data Generation: Data Pruning and Column Reordering

    cs.LG 2025-07 reject novelty 4.0 of 10

    PRRO combines signal-based data pruning and column reordering to improve the supervised learning utility of synthetic tabular data, but its evaluation is undermined by data manipulation and an ill-defined correlation measure.

Pith tools