Rapidata/117k_human_alignment_flux1.0_V_flux1.1Blueberry

Rapidata图像生成对齐数据集包含了117,000+全球范围内人类对AI生成图像的比较评估。数据集中的参与者针对给定的提示，对Flux 1.0和Flux 1.1Blueberry生成的两幅图像进行比较，并投票选择哪一幅更符合描述。数据集具有大规模、多样化的提示、两种领先模型之间的比较、严格的方法论以及丰富的参与者人口统计信息等特点。

The Rapidata Image Generation Alignment Dataset contains over 117,000 individual human preference votes from around the world on comparative evaluations of AI-generated images. Participants in the dataset are presented with a prompt and two images generated by Flux 1.0 and Flux 1.1Blueberry, respectively, and are asked to vote on which image better fits the description. The dataset is characterized by its massive scale, diverse prompts, comparisons between two leading models, rigorous methodology, and rich demographic information of annotators.