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

Kiwan Kwon Kangmin Kim Ho-Jin Lee Yeseong Jung Hyeongwoo Kong Vamsi K. Potluru Saerom Park Yongjae Lee

2026 · arXiv.org · Vol abs/2607.15606 · 1 citation · 30 references

Computer Science

TL;DR

This work presents a taxonomy-guided evaluation protocol for temporal fidelity, in which the applicable measurements are determined by the data rather than fixed in advance, and recasts utility and privacy evaluation over trajectories rather than isolated rows.

DOI Semantic Scholar

View source

Similar papers

#machine learning Preprint Aug 2026

Creation begins with understanding: LLMs as strategy designers for privacy-preserving tabular data synthesis

This work introduces Tabular Synthesis Strategy Designer, which uses an LLM to design synthesis procedures rather than directly generate records, and provides the LLM with tree-derived summaries of variable dependence rather than raw records, which produces Python programs for local execution and evaluation.

Jin-Meng Li, Quan Zhang, Hangting Ye et al. · 0 citations

Preprint Oct 2026

Soft Voting for Policy-Aware Private Data Synthesis

Blowfish privacy relaxes differential privacy (DP) by protecting only the attribute-value substitutions a data owner specifies as edges of a policy graph. A sparser policy can reduce the noise required by a mechanism, but only when the released statistic changes less across protected substitutions than across arbitrary...

Ying-Ge Hu, Gautham Ramesh Babu, Mostafa Milani · 0 citations

#artificial intelligence Preprint Sep 2026

Witness Overlap: Directional Provenance Inside Open-Weight Model Families

Witness Overlap is proposed, a prompt-free, training-free white-box test for directional provenance, and on 176 LLM checkpoints from 16 families, the one-witness test orients 95.3\% of parent-child decisions using Frobenius cosine.

Si-Yuan Li, Hao-Xuan Zeng, Xin Luo et al. · 0 citations

#artificial intelligence Review Sep 2026

FinancialAuditBench: Benchmark Construction under Differential Privacy Using Real-World Priors

FinancialAuditBench is introduced, a benchmark for evaluating agents on financial statement audit tasks, along with a framework for systematically generating synthetic engagements for model evaluation and training in privacy-sensitive domains.

Jerry Huang, Sarvesh Babu, Matt Van Buren et al. · 0 citations

#machine learning Preprint Oct 2026

Revealing After Overwriting: An Exponential POMDP OPE Lower Bound under History-Dependent Logging

Multi-step revealing can make off-policy evaluation tractable under memoryless logging. With history-dependent logging, state decodability and target-relevant evidence can separate. For every horizon $H\ge3$, we construct two exactly realizable POMDPs with four actions, at most four states per layer, a known logger, an...

You-Yu Luo, Peng-Zhan Zhou, Zhi-Da Qin et al. · 0 citations

#machine learning Preprint Sep 2026

Portable Causal Fairness Across Synthetic Data Generator Families

A new causal diffusion backbone yields the fairest release of any family the authors tested, at fidelity close to the marginals tier, and adding privacy guarantees don't make the data less fair.

Steven Golob, Sikha Pentyala, M. De Cock · 0 citations