Polypersona: Persona-Grounded LLM for Synthetic Survey Responses

TL;DR

This paper presents PolyPersona, a resource-efficient framework for generating persona-conditioned survey responses with small language models, enabling scalable, coherent, and bias-analyzable synthetic survey data across multiple domains.

Contribution

It introduces a novel, resource-efficient method for synthesizing multi-domain survey responses conditioned on personas using compact models and a dialogue-based data pipeline.

Findings

Small models like TinyLlama 1.1B achieve comparable performance to larger models.

The framework produces coherent, persona-aligned survey responses.

Synthetic data supports scalable evaluation and bias analysis.

Abstract

This paper introduces PolyPersona, a generative framework for synthesizing persona-conditioned survey responses across multiple domains. The framework instruction-tunes compact chat models using parameter-efficient LoRA adapters with 4-bit quantization under a resource-adaptive training setup. A dialogue-based data pipeline explicitly preserves persona cues, ensuring consistent behavioral alignment across generated responses. Using this pipeline, we construct a dataset of 3,568 synthetic survey responses spanning ten domains and 433 distinct personas, enabling controlled instruction tuning and systematic multi-domain evaluation. We evaluate the generated responses using a multi-metric evaluation suite that combines standard text generation metrics, including BLEU, ROUGE, and BERTScore, with survey-specific metrics designed to assess structural coherence, stylistic consistency, and sentiment alignment.Experimental results show that compact models such as TinyLlama 1.1B and Phi-2 achieve performance comparable to larger 7B to 8B baselines, with a highest BLEU score of 0.090 and ROUGE-1 of 0.429. These findings demonstrate that persona-conditioned fine-tuning enables small language models to generate reliable and coherent synthetic survey data. The proposed framework provides an efficient and reproducible approach for survey data generation, supporting scalable evaluation while facilitating bias analysis through transparent and open protocols.