SYNTHONY: A Stress-Aware, Intent-Conditioned Agent for Deep Tabular Generative Models Selection

Hochan Son; Xiaofeng Lin; Jason Ni; Guang Cheng

cs.LGMar 31, 2026

SYNTHONY: A Stress-Aware, Intent-Conditioned Agent for Deep Tabular Generative Models Selection

Authors: Hochan Son, Xiaofeng Lin, Jason Ni, Guang Cheng

Organizations: Department of Statistics, University of California, Los Angeles · Department of Mathematics, University of California, Los Angeles

Abstract

Deep generative models for tabular data (GANs, diffusion models, and LLM-based generators) exhibit highly non-uniform behavior across datasets; the best-performing synthesizer family depends strongly on distributional stressors such as long-tailed marginals, high-cardinality categorical, Zipfian imbalance, and small-sample regimes. This brittleness makes practical deployment challenging, especially when users must balance competing objectives of fidelity, privacy, and utility. We study {intent-conditioned tabular synthesis selection}: given a dataset and a user intent expressed as a preference over evaluation metrics, the goal is to select a synthesizer that minimizes regret relative to an intent-specific oracle. We propose {stress profiling}, a synthesis-specific meta-feature representation that quantifies dataset difficulty along four interpretable stress dimensions, and integrate it into {SYNTHONY}, a selection framework that matches stress profiles against a calibrated capability registry of synthesizer families. Across a benchmark of 7 datasets, 10 synthesizers, and 3 intents, we demonstrate that stress-based meta-features are highly predictive of synthesizer performance: a $k$ NN selector using these features achieves strong Top-1 selection accuracy, substantially outperforming zero-shot LLM selectors and random baselines. We analyze the gap between meta-feature-based and capability-based selection, identifying the hand-crafted capability registry as the primary bottleneck and motivating learned capability representations as a direction for future work.

Figures & tables

Strategy	Top-1	Top-3	Spear.	NDCG
$k$ NN (k=3) † (oracle-informed)	.556	.778	.578	.918
Random (E[1000])	.111	.304	$-.001$	.815
Vanilla LLM (GPT-4o-mini)	.000	.111	$-.087$	.784
Static Heuristic	.333	.778	.422	.889
Synthony (joint opt)	.000	.556	.557	.893
Synthony (SF-only)	.444	.667	.573	.914

Table 1: Test set performance of intent-conditioned model selection strategies (9 dataset-intent pairs). Primary metrics are italicized .

k

NN † uses oracle neighbor rankings from training datasets and serves as an oracle-informed reference point , not a deployable zero-knowledge method. Bold indicates best among zero-knowledge methods. With 9 test pairs, each Top-1 prediction shifts accuracy by

{\approx}

11.1 percentage points.

Variant	Top-1	Top-3	Spear.	NDCG
Vanilla LLM (no Synthony )	.000	.111	$-.087$	.784
$-$ Stress $-$ Focus (bare scoring)	.444	.667	.478	.904
$-$ Focus scaling (stress only)	.444	.667	.602	.910
$-$ Stress profiling (focus only)	.333	.667	.430	.901
Full Synthony (SF-only)	.444	.667	.573	.914

Table 2: Ablation study on the SF-only pipeline (test set, 9 pairs). Components: (A) stress profiling, (B) capability matching, (C) focus scaling. Stress profiling provides the strongest signal; adding focus scaling degrades test performance due to overfitting at this sample size (

n{=}12

training pairs, 18 parameters).

Predictor	Top-1	Top-3	Spear.	NDCG
Majority vote	.333	.778	.300	.899
Logistic regression	.556	.889	.293	.893
Decision tree	.222	.667	.299	.900
$k$ NN (k=3)	.556	.778	.578	.918

Table 3: Predicting the best synthesizer from dataset meta-features (test set, 9 pairs). Majority vote’s

33.3\%

Top-1 reflects CART’s frequent wins (8/21 oracle-best); the discriminative signal is in Spearman, where

k

NN (

0.578

) substantially outperforms frequency-based and parametric baselines (

0.293

–

0.300

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Split	Top-1	Top-3	Spearman	NDCG
Training (12 pairs)	.250	.750	.610	.897
Test (9 pairs)	.000	.556	.557	.893

Appendix

Table 4: Joint optimization results on train and test splits.

	Skew		Cardinal.		Zipfian		Small data		Correl.		Privacy
Model	Orig	Opt	Orig	Opt	Orig	Opt	Orig	Opt	Orig	Opt	Orig	Opt
AIM	3	2	0	0	1	3	2	4	3	0	4	2
ARF	2	4	4	2	3	4	4	0	4	1	0	1
AutoDiff	1	4	3	1	2	3	2	0	1	3	0	4
BayesianNet.	3	4	4	4	2	2	4	4	3	2	0	4
CART	3	4	4	2	2	2	4	4	4	3	0	4

Appendix

Table 5: Capability scores: hand-crafted (left) vs. optimized (right). Changes of

\geq 2

are bolded.

Intent	Skew	Cardinal.	Zipfian	Small	Correl.	Privacy
Privacy	5.69	3.78	7.32	9.83	8.90	9.10
Fidelity	2.43	7.36	4.10	3.11	2.78	9.44
Utility	0.68	2.59	3.71	5.36	1.42	2.96

Appendix

Table 6: Optimized scale factors

\alpha_{i,j}

per intent and capability dimension.

Panel of Figure 1 — Figure 1: SYNTHONY system architecture. Solid arrows denote the real-time recommendation pipeline; dashed arrows denote the offline calibration loop (Section 3.4 ). Data flows left through profiling, center through scoring, and returns right through the API.

Panel of Figure 2 — Figure 2: Rule-based scoring pipeline. The match function $m_{j}$ implements a decay curve: 1.0 (exact match), 0.7 (one level below), 0.4 (two levels below), 0.0 (otherwise). Scale factors $\alpha_{i,j}$ are applied when an intent is specified. The hard-problem path is bypassed when scale factors are provided to allow the optimizer full control.

Condition	Match Score	Interpretation
$c_{g,j}\geq r_{j}$	1.0	Capability meets or exceeds requirement
$c_{g,j}=r_{j}-1$	0.7	Near miss (one level below)
$c_{g,j}=r_{j}-2$	0.4	Moderate gap (two levels below)
$c_{g,j}<r_{j}-2$	0.0	Insufficient capability

Appendix

Table 7: Score decay curve: match score as a function of capability gap.

Panel of Figure 3 — Figure 3: API architecture with SQLite persistence. Double arrows indicate read/write paths. Uploaded files are stored on disk; profiles and recommendations are cached in SQLite for session-based retrieval.

Table	Key Columns	Relationships	Purpose
sessions	session_id (PK), ip_address , expires_at	$\to$ datasets (1:N)	Track client sessions (30-day TTL)
datasets	dataset_id (PK), session_id (FK), file_path	$\to$ analyses (1:N)	Uploaded file metadata
analyses	analysis_id (PK), dataset_id (FK), profile_json	$\to$ system_prompts	Cached profiles + recommendations
system_prompts	prompt_id (PK), version , is_active	—	Versioned LLM system prompts
audit_logs	log_id (PK), session_id , action	—	Request audit trail

Appendix

Table 8: SQLite database schema. PK = primary key, FK = foreign key.

Method	Endpoint	Description
POST	/analyze	Upload CSV/Parquet $\to$ stress profile + column analysis (cached)
POST	/recommend	Profile $\to$ ranked recommendation (rule/LLM/hybrid)
POST	/analyze-and-recommend	One-shot: upload $\to$ profile $\to$ recommendation
GET	/models	List models with optional filters (type, CPU, DP)
GET	/models/{name}	Detailed model capabilities and constraints
GET	/health	Component status (analyzer, recommender, LLM availability)

Appendix

Table 9: REST API endpoint summary.

Panel of Figure 4 — Figure 4: MCP server architecture. The server exposes SYNTHONY’s capabilities through three MCP primitives: Tools (executable functions), Resources (read-only data), and Prompts (guided workflows). Communication uses JSON-RPC 2.0 over stdio.

Directory / File	Contents	Paper Reference
src/synthony/core/	Data loaders, StochasticDataAnalyzer , schemas	Section 3.2
src/synthony/detectors/	Skewness, cardinality, Zipfian, correlation detectors	Section 3.2
src/synthony/recommender/	ModelRecommendationEngine , scoring logic	Section 3.3
config/model_capabilities.json	Capability registry (v7.0.0, 15 models $\times$ 6 dims)	Section 3.3
scripts/optimize_scaling.py	Bayesian optimization of scale factors (Optuna/TPE)	Section A
src/synthony/api/	FastAPI REST server with SQLite persistence	Appendix C.3

Appendix

Table 10: Repository structure and correspondence to paper sections.

Synthetic Data Selection

Synthetic Data Selection studies how to choose generated examples that improve downstream model training while limiting noise, redundancy, and cost.

See all 31 papers

ReTabSyn: Realistic Tabular Data Synthesis via Reinforcement Learning
Mar 11, 2026Xiaofeng Lin, Seungbae Kim, Zhuoya Li +3Synthetic Tabular Data Generation Tabular Data Synthesis
University of California, Los Angeles, USA · University of South Florida, USA · Cisco, USA +1
Creation begins with understanding: LLMs as strategy designers for privacy-preserving tabular data synthesis
Aug 30, 2026Jinmeng Li, Quan Zhang, Hangting Ye +4Synthetic Tabular Data Generation Privacy-Preserving ML
School of Artificial Intelligence, Jilin University, Changchun, China · Broad College of Business, Michigan State University, East Lansing, USA · Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia +2
TabQueryBench: A Query-Centric Benchmark for Synthetic Tabular Data
Jul 4, 2026Jialin Zhang, Fenghao Dong, Yajie Zhou +2Benchmark Design Synthetic Tabular Data Generation
University of Hong Kong · Tongji University · Carnegie Mellon University +1

Mar 11, 2026J/K move · Enter open · S save

ReTabSyn: Realistic Tabular Data Synthesis via Reinforcement Learning

Xiaofeng Lin, Seungbae Kim, Zhuoya Li, Zachary DeSoto+2

University of California, Los Angeles, USA · University of South Florida, USA · Cisco, USA+1

Synthetic Tabular Data Generation Tabular Data Synthesis Conditional Generative Modeling RL for Generative Models

Open paper AskPDF arXiv

Deep generative models can help with data scarcity and privacy by producing synthetic training data, but they struggle in low-data, imbalanced tabular settings to fully learn the complex data distribution. We argue that striving for the full joint distribution could be overkill; for greater data efficiency, models should prioritize learning the conditional distribution

P(y\mid \bm{X})

, as suggested by recent theoretical analysis. Therefore, we overcome this limitation with ReTabSyn, a Reinforced Tabular Synthesis pipeline that provides direct feedback on feature correlation preservation during synthesizer training. This objective encourages the generator to prioritize the most useful predictive signals when training data is limited, thereby strengthening downstream model utility. We empirically fine-tune a language model-based generator using this approach, and across benchmarks with small sample sizes, class imbalance, and distribution shift, ReTabSyn consistently outperforms state-of-the-art baselines. Moreover, our approach can be readily extended to control various aspects of synthetic tabular data, such as applying expert-specified constraints on generated observations.

Abstract

Synthetic Data Selection

Explore similar work

ReTabSyn: Realistic Tabular Data Synthesis via Reinforcement Learning

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

TabQueryBench: A Query-Centric Benchmark for Synthetic Tabular Data