enkryptai/txt-image-bias-dataset

txt-image-bias-dataset是一个包含基于潜在社会偏见（如宗教、种族和性别）分类的文本提示的数据集。该数据集旨在帮助研究文本到图像模型中的偏见缓解，通过识别可能导致生成图像中出现偏见或刻板印象的提示。数据集包含两个列：提示和与之关联的偏见类别，包括宗教、种族和性别。共有60个数据点，平均分布在三个类别中。

The txt-image-bias-dataset is a collection of text prompts categorized based on potential societal biases related to religion, race, and gender. The dataset is designed to aid research on bias mitigation in text-to-image models by identifying prompts that may lead to biased or stereotypical representations in generated images. It includes two columns: prompt and category, with the categories being Religion, Race, and Gender. There are a total of sixty data points, evenly distributed across the three categories.