Rapidata/HunyuanImage-2.1_t2i_human_preference

Rapidata Hunyuan Image 2.1偏好数据集包含约400,000个来自约50,000名注释者的响应，用于评估Hunyuan Image 2.1模型在偏好、一致性和对齐方面的表现。该数据集通过Rapidata Python API在7小时内收集，并与其他18个模型进行比较。数据集包含训练集和测试集，以及相关的元数据，如prompt、image1、image2、model1、model2等。数据集支持多种任务类别，如文本到图像、图像到文本、图像分类和强化学习。

The Rapidata Hunyuan Image 2.1 Preference dataset includes approximately 400,000 responses from about 50,000 annotators, used to evaluate the Hunyuan Image 2.1 model in terms of preference, coherence, and alignment. The dataset was collected in less than 7 hours using the Rapidata Python API and compared with 18 other models. It includes training and test sets, as well as related metadata such as prompt, image1, image2, model1, model2, etc. The dataset supports various task categories such as text-to-image, image-to-text, image classification, and reinforcement learning.