CuRe：文本到图像系统长尾中的文化差距
CuRe: Cultural Gaps in the Long Tail of Text-to-Image Systems
AI总结：
本文提出 CuRe 文化代表性基准与评分套件，通过属性规范边际效用衡量 T2I 系统的全球南方文化偏差，并在多种模型上验证其与人类判断的相关性。
摘要：
流行的文本到图像（T2I）系统是在网络抓取数据上训练的，这些数据严重以美国和欧洲为中心，对全球南方文化的代表性不足。为分析这些偏差，我们提出 CuRe，这是一个新颖且可扩展的文化代表性基准测试与评分套件，它利用属性规范对 T2I 系统的边际效用作为人类判断的代理指标。我们的 CuRe 基准数据集具有基于众包 Wikimedia 知识图谱构建的新颖分类层级，包含 300 件文化产物，横跨 32 个文化子类别，并归入六大文化轴（食物、艺术、时尚、建筑、庆典和人物）。数据集的分类层级使 CuRe 评分器能够通过分析 T2I 系统对文本条件信息量增加的响应来评估这些系统，从而实现细粒度文化比较。我们通过实验观察到，在各类图像编码器（SigLIP 2、AIMV2 和 DINOv2）、视觉-语言模型（OpenCLIP、SigLIP 2、Gemini 2.0 Flash）以及最先进的文本到图像系统（包括 Stable Diffusion 的三个变体 1.5、XL、3.5 Large，以及 FLUX.1 [dev]、Ideogram 2.0 和 DALL-E 3）上，我们的这类评分器与人类关于感知相似性、图文对齐和文化多样性的判断具有更强的相关性。代码和数据集已开源，可在 https://aniketrege.github.io/cure/ 获取。
英文摘要：
Popular text-to-image (T2I) systems are trained on web-scraped data, which is heavily Amero and Euro-centric, underrepresenting the cultures of the Global South. To analyze these biases, we introduce CuRe, a novel and scalable benchmarking and scoring suite for cultural representativeness that leverages the marginal utility of attribute specification to T2I systems as a proxy for human judgments. Our CuRe benchmark dataset has a novel categorical hierarchy built from the crowdsourced Wikimedia knowledge graph, with 300 cultural artifacts across 32 cultural subcategories grouped into six broad cultural axes (food, art, fashion, architecture, celebrations, and people). Our dataset's categorical hierarchy enables CuRe scorers to evaluate T2I systems by analyzing their response to increasing the informativeness of text conditioning, enabling fine-grained cultural comparisons. We empirically observe much stronger correlations of our class of scorers to human judgments of perceptual similarity, image-text alignment, and cultural diversity across image encoders (SigLIP 2, AIMV2 and DINOv2), vision-language models (OpenCLIP, SigLIP 2, Gemini 2.0 Flash) and state-of-the-art text-to-image systems, including three variants of Stable Diffusion (1.5, XL, 3.5 Large), FLUX.1 [dev], Ideogram 2.0, and DALL-E 3. The code and dataset is open-sourced and available at https://aniketrege.github.io/cure/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
