adi-gulati/LookismBench-SyntheticFaces

Summary LookismBench-Faces is a large-scale synthetic dataset designed to study algorithmic lookism in text-to-image (T2I) generative models and downstream computer vision systems. The dataset consists of 26,400 synthetic human faces generated using Stable Diffusion 2.1 and 3.5 Medium under controlled prompt conditions. It enables systematic analysis of how generative AI models encode associations between facial attractiveness, semantic attributes, and gender classification… See the full description on the dataset page: https://huggingface.co/datasets/adi-gulati/LookismBench-SyntheticFaces.