Seq2Synth: Benchmarking Temporal Fidelity in Synthetic Sequential Tabular Data
Abstract
Synthetic sequential tabular data are increasingly used for privacy-preserving data sharing and research, yet conventional tabular metrics often overlook temporal structure. Existing single-table and relational evaluation protocols largely collapse records into static distributions, leaving key temporal properties insufficiently evaluated. We introduce Seq2Synth, a unified benchmark for assessing these properties. Its taxonomy characterizes temporal and schema properties to determine applicable evaluations, covering timestamp, cross-sectional, longitudinal, and structural fidelity, alongside trajectory-aware utility and privacy. Across seven core datasets from a 13-dataset benchmark and eight generators, models with near-perfect static fidelity still violate basic temporal constraints, producing duplicate timestamps, irregular intervals, and incomplete observation grids. Moreover, static and temporal-aware rankings diverge substantially, showing that temporal fidelity must be evaluated directly rather than inferred from static or relational scores. Project page and online appendices are available at: https://seq2synth.github.io/.
Versions
Related
- Cost-Aware Iteration Allocation for Sequential Multi-Fidelity First-Order Optimization
- Beyond Distributional Fidelity: Causal-Penalized Diffusion for Synthetic Tabular Data
- Reproducible benchmarking of four machine-learning classifiers on a large synthetic heart-disease competition dataset
- A mechanism-guided simulation framework for synthetic tax-risk data: Constraint validation and reproducible benchmarking
- Self-Supervised Tuning of Low-Fidelity Optimization Models via High-Fidelity Feedback
BibTeX
@misc{kwon2026seq2synth,
title = {{Seq2Synth: Benchmarking Temporal Fidelity in Synthetic Sequential Tabular Data}},
author = {Kiwan Kwon and Kangmin Kim and Hojin Lee and Yeseong Jung and Hyeongwoo Kong and Vamsi K. Potluru and Saerom Park and Yongjae Lee},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2607.15606},
eprint = {2607.15606},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2607.15606},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Seq2Synth: Benchmarking Temporal Fidelity in Synthetic Sequential Tabular Data
AU - Kiwan Kwon
AU - Kangmin Kim
AU - Hojin Lee
AU - Yeseong Jung
AU - Hyeongwoo Kong
AU - Vamsi K. Potluru
AU - Saerom Park
AU - Yongjae Lee
PY - 2026
DA - 2026/08/31
PB - arXiv
DO - 10.48550/arxiv.2607.15606
UR - https://arxiv.org/abs/2607.15606
AB - Synthetic sequential tabular data are increasingly used for privacy-preserving data sharing and research, yet conventional tabular metrics often overlook temporal structure. Existing single-table and relational evaluation protocols largely collapse records into static distributions, leaving key temporal properties insufficiently evaluated. We introduce Seq2Synth, a unified benchmark for assessing these properties. Its taxonomy characterizes temporal and schema properties to determine applicable evaluations, covering timestamp, cross-sectional, longitudinal, and structural fidelity, alongside trajectory-aware utility and privacy. Across seven core datasets from a 13-dataset benchmark and eight generators, models with near-perfect static fidelity still violate basic temporal constraints, producing duplicate timestamps, irregular intervals, and incomplete observation grids. Moreover, static and temporal-aware rankings diverge substantially, showing that temporal fidelity must be evaluated directly rather than inferred from static or relational scores. Project page and online appendices are available at: https://seq2synth.github.io/.
ER -