Rapidata/Seedream-3_t2i_human_preference

Rapidata Seedream 3 Preference数据集包含约30,000名标注者提供的超过40万个人类反馈，用于评估OpenAI 4o模型在偏好、一致性和对齐性三个方面的表现。数据集包含提示、图像、模型名称以及与评估标准相关的各种分数。数据集分为训练集，并具有指定的大小和下载大小。该数据集受CDLA-Permissive-2.0许可协议的保护，并被归类为多个任务类别和大小类别。数据集描述为文本到图像、图像到文本、图像分类和强化学习数据集，重点关注人类偏好、一致性和对齐性。数据集旨在评估Seedream 3、Ideogram V2、Recraft V2、Lumina等模型在多个评估标准方面的表现。

The Rapidata Seedream 3 Preference dataset consists of over 400,000 human responses from approximately 30,000 annotators, used to evaluate the OpenAI 4o model across three categories: preference, coherence, and alignment. The dataset features include prompts, images, model names, and various scores related to the evaluation criteria. The dataset is split into a training set and has a specified size and download size. It is licensed under the CDLA-Permissive-2.0 license and is categorized under several task categories and size categories. The dataset is described as a text-to-image, image-to-text, image-classification, and reinforcement-learning dataset, with a focus on human preferences, coherence, and alignment. The dataset is designed to evaluate models such as Seedream 3, Ideogram V2, Recraft V2, Lumina, and others, across various evaluation criteria.