mlsanityData to 7 Oct 2026

Beyond the Stitching Assumption: A Unified Framework for Multimodal Synthetic Data Evaluation via Semantic Quantization

arXiv 2609.22149 ↗First posted 26 Aug 2026cs.CL cs.LGcited by 0

On the map: Synthetic data and self-improvement for LLMs → Privacy-preserving tabular data synthesis

Abstract (arXiv)

Multimodal synthetic datasets combine structured attributes with free text, but are often evaluated separately. Such metrics can remain high after tabular--text pairings are disrupted. We present a projection-based evaluator for tabular--text synthetic data. A fixed sentence encoder maps text to embeddings, (k)-means converts them to cluster states, and tabular variables are represented as categorical or quantile-binned states. Real and synthetic contingency tables are compared using Jensen--Shannon divergence (JSD), normalized mutual information (NMI), conditional JSD (cJSD), and joint-state entropy. We also report text-to-attribute (T2A) utility and a holdout-calibrated proximity flag rate (PFR) as a representation-level diagnostic. A text-permutation control preserves both marginal distributions while disrupting their pairing. Experiments on Amazon Reviews, Kiva Loans, and the Employment Scam Aegean Dataset show that modality-specific scores remain high under this control. The projection diagnostics detect disruption when real projected dependence exceeds a permutation baseline, but are less informative for weak or sparse projections. Some conditioned LLM baselines also exhibit stronger measured dependence than the corresponding real-data projections. These results support explicit cross-modal evaluation with permutation baselines and coverage reporting.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Privacy-preserving tabular data synthesis, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 18 of its 38 references that are in the map. First posted on arXiv 26 Aug 2026.

Builds on

18 of 38 references.

Closest ancestors · 4

Same direction, newest first

  1. 2024A Multi-Faceted Evaluation Framework for Assessing Synthetic Data Generated by Large Language Models20 Apr
  2. 2024EPIC: Effective Prompting for Imbalanced-Class Data Synthesis in Tabular Data Classification via Large Language Models15 Apr
  3. 2023TabuLa: Harnessing Language Models for Tabular Data Synthesis19 Oct
  4. 2023Mixed-Type Tabular Data Synthesis with Score-based Diffusion in Latent Space14 Oct

Other references · 14

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  3. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  4. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,462
  5. 2017Multimodal Machine Learning: A Survey and Taxonomycited by 1,004
  6. 2019Modeling Tabular data using Conditional GANcited by 695
  7. 2019ClinicalBERT: Modeling Clinical Notes and Predicting Hospital Readmissioncited by 487
  8. 2021MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontierscited by 466
  9. 2022TabDDPM: Modelling Tabular Data with Diffusion Modelscited by 373
  10. 2022Language Models are Realistic Tabular Data Generatorscited by 290
  11. 2022TabLLM: Few-shot Classification of Tabular Data with Large Language Modelscited by 273
  12. 2023Synthcity: facilitating innovative use cases of synthetic data in different data modalitiescited by 76
2 more
  1. 2020A Tutorial on Learning With Bayesian Networkscited by 56
  2. 2025MissDDIM: Deterministic and Efficient Conditional Diffusion for Tabular Data Imputationcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Understanding the Influence of Synthetic Data for Text Embedders
  2. 2025An Empirical Study of Validating Synthetic Data for Text-Based Person Retrieval
  3. 2026Mind the Gap? A Distributional Comparison of Real and Synthetic Priors for Tabular Foundation Models
  4. 2025mmE5: Improving Multimodal Multilingual Embeddings via High-quality Synthetic Data
  5. 2024A Synthetic Dataset for Personal Attribute Inference
  6. 2025Synthetic Data Can Mislead Evaluations: Membership Inference as Machine Text Detection
  7. 2026Improving TabPFN's Synthetic Data Generation by Integrating Causal Structure
  8. 2025Privacy-Preserving Synthetic Review Generation with Diverse Writing Styles Using LLMs