shirsho12/text2image-10k-with-spectacles-pairs

这是一个包含10,000条文本信息的文本数据集，其中9,000条提示来自jackyhate/text-to-image-2M数据集，另外1,000条是用户提供的带有眼镜的配对信息。数据集分为训练集和测试集，特征包括文本内容、是否来自配对、是否包含戴眼镜的短语以及数据来源。该数据集旨在用于图像生成模型中的偏差/操纵分析。

This is a text dataset containing 10,000 entries, including 9,000 prompts from the jackyhate/text-to-image-2M dataset and 1,000 pairs of user-provided information with glasses. The dataset is split into training and test sets, with features including text content, whether it is from a pair, whether it includes a phrase about wearing glasses, and the data source. The dataset is intended for bias/steering analyses in image generation models.