LLMSynthor: Macro-Aligned Micro-Records Synthesis with Large Language Models

TL;DR

LLMSynthor leverages large language models to generate realistic micro-level data aligned with macro-statistics, enabling credible simulations in social sciences and urban studies without extensive data collection.

Contribution

This work introduces a macro-aware LLM-based framework that iteratively synthesizes micro-records matching target macro-statistics, using a nonparametric copula approach and proposal sampling for efficiency.

Findings

Achieves high realism and statistical fidelity in synthetic data across multiple domains.

Effectively captures joint dependencies among variables in generated micro-records.

Demonstrates broad applicability to economics, social science, and urban studies.

Abstract

Macro-aligned micro-records are crucial for credible simulations in social science and urban studies. For example, epidemic models are only reliable when individual-level mobility and contacts mirror real behavior, while aggregates match real-world statistics like case counts or travel flows. However, collecting such fine-grained data at scale is impractical, leaving researchers with only macro-level data. LLMSynthor addresses this by turning a pretrained LLM into a macro-aware simulator that generates realistic micro-records consistent with target macro-statistics. It iteratively builds synthetic datasets: in each step, the LLM generates batches of records to minimize discrepancies between synthetic and target aggregates. Treating the LLM as a nonparametric copula allows the model to capture realistic joint dependencies among variables. To improve efficiency, LLM Proposal Sampling guides the LLM to propose targeted record batches, specifying variable ranges and counts, to efficiently correct discrepancies while preserving realism grounded in the model's priors. Evaluations across domains (mobility, e-commerce, population) show that LLMSynthor achieves strong realism, statistical fidelity, and practical utility, making it broadly applicable to economics, social science, and urban studies.