分析160个主题下文本生成图像的偏见，发现模型放大社会刻板印象。
Hidden Bias in the Machine: Stereotypes in Text-to-Image Models
- 构建160个主题的多角度提示集，生成超1.6万张图像
- 对比发现性别、种族等特征呈现显著不平等分布
- 揭示模型对社会偏见的复制与强化，适合关注AI伦理者
文本到图像（T2I）模型已革新视觉内容生成，可从自然语言提示生成高度逼真的图像。然而，其可能复制并放大现有社会偏见的问题仍引发关注。为此，我们构建了涵盖职业、特质、行为、意识形态、情绪、家庭角色、场所描述、宗教信仰及人生事件等主题的多样化提示集，共160个独特主题，每个主题设计多种提示变体以覆盖多元语义与视角。使用基于UNet的Stable Diffusion 1.5和基于DiT的Flux-1模型，采用原始检查点，在一致设置下生成超过16,000张图像，并额外收集8,000张来自Google Image Search的对比图像。所有输出经筛选，剔除抽象、扭曲或无意义结果。分析显示，生成图像在性别、种族、年龄、体型等人类相关特征上存在显著差异，这些差异常与社会叙事中的有害刻板印象一致。我们讨论了这些发现的影响，并强调需构建更包容的数据集与开发实践，以推动生成式视觉系统的公平性。
原文摘要 · Abstract (English)
Text-to-Image (T2I) models have transformed visual content creation, producing highly realistic images from natural language prompts. However, concerns persist around their potential to replicate and magnify existing societal biases. To investigate these issues, we curated a diverse set of prompts spanning thematic categories such as occupations, traits, actions, ideologies, emotions, family roles, place descriptions, spirituality, and life events. For each of the 160 unique topics, we crafted multiple prompt variations to reflect a wide range of meanings and perspectives. Using Stable Diffusion 1.5 (UNet-based) and Flux-1 (DiT-based) models with original checkpoints, we generated over 16,000 images under consistent settings. Additionally, we collected 8,000 comparison images from Google Image Search. All outputs were filtered to exclude abstract, distorted, or nonsensical results. Our analysis reveals significant disparities in the representation of gender, race, age, somatotype, and other human-centric factors across generated images. These disparities often mirror and reinforce harmful stereotypes embedded in societal narratives. We discuss the implications of these findings and emphasize the need for more inclusive datasets and development practices to foster fairness in generative visual systems.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。