Zhaorun/MJ-Bench-Image
收藏官方服务：
资源简介：
MJ-Bench-Image数据集包含了用于评估文本到图像生成中多模态评判员的MJ-Bench基准中的图像对。数据集分为多个类别，包括对齐、偏见、组成、质量、安全性和可视化等，每个样本包含文本提示、正面图像、负面图像、偏好标签、评估类别、被评估的具体方面以及使用的生成模型。
The MJ-Bench-Image Dataset contains image pairs from the MJ-Bench benchmark for evaluating multimodal judges in text-to-image generation. It is organized into categories such as Alignment, Bias, Composition, Quality, Safety, and Visualization, with each sample including a text prompt, a positive image, a negative image, a preference label, an evaluation category, a specific aspect being evaluated, and the generation model used.
提供机构：
Zhaorun
