Aligned but Stereotypical? The Hidden Influence of System Prompts on Social Bias in LVLM-Based Text-to-Image Models

TL;DR

This paper investigates how system prompts influence social bias in LVLM-based text-to-image models, revealing prompts as a key factor in bias propagation and proposing a framework to mitigate this bias without retraining.

Contribution

It identifies system prompts as a primary source of bias in LVLM-based T2I models and introduces FairPro, a framework to reduce bias through self-auditing prompts at test time.

Findings

LVLM-based models produce more social bias than non-LVLM models.

System prompts encode demographic priors that influence bias in generated images.

FairPro significantly reduces demographic bias while maintaining image quality.

Abstract

Large vision-language model (LVLM) based text-to-image (T2I) systems have become the dominant paradigm in image generation, yet whether they amplify social biases remains insufficiently understood. In this paper, we show that LVLM-based models produce markedly more socially biased images than non-LVLM-based models. We introduce a 1,024 prompt benchmark spanning four levels of linguistic complexity and evaluate demographic bias across multiple attributes in a systematic manner. Our analysis identifies system prompts, the predefined instructions guiding LVLMs, as a primary driver of biased behavior. Through decoded intermediate representations, token-probability diagnostics, and embedding-association analyses, we reveal how system prompts encode demographic priors that propagate into image synthesis. To this end, we propose FairPro, a training-free meta-prompting framework that enables LVLMs to self-audit and construct fairness-aware system prompts at test time. Experiments on two LVLM-based T2I models, SANA and Qwen-Image, show that FairPro substantially reduces demographic bias while preserving text-image alignment. We believe our findings provide deeper insight into the central role of system prompts in bias propagation and offer a practical, deployable approach for building more socially responsible T2I systems.