Saved in:
Bibliographic Details
Main Authors: Zhao, Jianpeng, Yuan, Chenyu, Luo, Weiming, Xie, Haoling, Zhang, Guangwei, Quan, Steven Jige, Yuan, Zixuan, Wang, Pengyang, Zhang, Denghui
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.06337
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911624532590592
author Zhao, Jianpeng
Yuan, Chenyu
Luo, Weiming
Xie, Haoling
Zhang, Guangwei
Quan, Steven Jige
Yuan, Zixuan
Wang, Pengyang
Zhang, Denghui
author_facet Zhao, Jianpeng
Yuan, Chenyu
Luo, Weiming
Xie, Haoling
Zhang, Guangwei
Quan, Steven Jige
Yuan, Zixuan
Wang, Pengyang
Zhang, Denghui
contents Questionnaire-based surveys are foundational to social science research and public policymaking, yet traditional survey methods remain costly, time-consuming, and often limited in scale. Although prior work has explored large language models (LLMs) as virtual survey respondents, existing studies often address narrow task settings, focus on single sociological domains, or lack a unified evaluation framework that enables systematic comparison across diverse datasets and models. To address these gaps, we introduce two complementary task abstractions: Partial Attribute Simulation (PAS), where LLMs predict missing attributes from incomplete respondent profiles, and Full Attribute Simulation (FAS), where LLMs generate complete synthetic datasets under zero-context and context-enhanced conditions. Both are framed as diagnostic and exploratory tools rather than replacements for human data collection. We curate LLM-S^3 (Large Language Model-based Sociodemographic Survey Simulation), a benchmark spanning 11 real-world public datasets across four sociological domains, and evaluate GPT-3.5/4 Turbo and LLaMA 3.0/3.1-8B under zero-shot and few-shot settings. Our evaluation reveals consistent performance trends across model families, highlights failure modes in structured output generation, and demonstrates how context and prompt design affect simulation fidelity. Our code and dataset are available at: https://github.com/dart-lab-research/LLM-S-Cube-Benchmark
format Preprint
id arxiv_https___arxiv_org_abs_2509_06337
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Large Language Models as Virtual Survey Respondents: Evaluating Sociodemographic Response Generation
Zhao, Jianpeng
Yuan, Chenyu
Luo, Weiming
Xie, Haoling
Zhang, Guangwei
Quan, Steven Jige
Yuan, Zixuan
Wang, Pengyang
Zhang, Denghui
Artificial Intelligence
Questionnaire-based surveys are foundational to social science research and public policymaking, yet traditional survey methods remain costly, time-consuming, and often limited in scale. Although prior work has explored large language models (LLMs) as virtual survey respondents, existing studies often address narrow task settings, focus on single sociological domains, or lack a unified evaluation framework that enables systematic comparison across diverse datasets and models. To address these gaps, we introduce two complementary task abstractions: Partial Attribute Simulation (PAS), where LLMs predict missing attributes from incomplete respondent profiles, and Full Attribute Simulation (FAS), where LLMs generate complete synthetic datasets under zero-context and context-enhanced conditions. Both are framed as diagnostic and exploratory tools rather than replacements for human data collection. We curate LLM-S^3 (Large Language Model-based Sociodemographic Survey Simulation), a benchmark spanning 11 real-world public datasets across four sociological domains, and evaluate GPT-3.5/4 Turbo and LLaMA 3.0/3.1-8B under zero-shot and few-shot settings. Our evaluation reveals consistent performance trends across model families, highlights failure modes in structured output generation, and demonstrates how context and prompt design affect simulation fidelity. Our code and dataset are available at: https://github.com/dart-lab-research/LLM-S-Cube-Benchmark
title Large Language Models as Virtual Survey Respondents: Evaluating Sociodemographic Response Generation
topic Artificial Intelligence
url https://arxiv.org/abs/2509.06337