Rich-Human-Feedback Image Dataset

Rich-Human-Feedback is a dataset containing rich human feedback for text-to-image generation, released by the Rapidata organization in 2025, which aims to provide rich feedback for the training and evaluation of text-to-image generation models.Rich Human Feedback for Text-to-Image GenerationThe dataset contains 15k images and 1.5 million annotations from over 150,000 people, including feedback such as image ratings, semantic consistency, and correction suggestions. Through multi-dimensional annotation, we can obtain image ratings in terms of style, coherence, consistency with prompt words, and mark misaligned words and image problem areas, helping developers to deeply understand the advantages and limitations of the model.