OpenSafetyLab/t2i_safety_dataset

T2ISafety是一个用于评估文本到图像模型在毒性、公平性和隐私性方面的综合安全基准。该数据集包含70K精心收集的提示构建的12个任务和44个子任务，以及68K人工注释的图像，用于训练和测试T2I安全评估器。数据集展示了公平性、毒性和隐私性三个主要领域的提示-图像对。T2ISafety is a comprehensive safety benchmark designed to evaluate Text-to-Image (T2I) models across three key domains: toxicity, fairness, and privacy. The dataset includes 70K prompts meticulously collected to build 12 tasks and 44 sub-tasks, along with 68K manually annotated images for training and testing T2I safety evaluators, showcasing prompt-image pairs across the three main domains of fairness, toxicity, and privacy.