A large-scale crowdsourced dataset examining how beauty filters influence perceptions of the attractiveness halo effect. Collected from 2,748 participants rating 472 individuals with and without cosmetic filters applied, across multiple trait dimensions including intelligence, competence, and trustworthiness.

A benchmark dataset of approximately 26,000 synthetic face images generated with Stable Diffusion 2.1 and 3.5, designed to evaluate algorithmic bias in appearance-based discrimination across text-to-image generation and classification models. Enables systematic assessment of lookism in generative AI systems.