Reproduce “Investigating Disability Representations in Text-to-Image Models” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yang Tian, Yu Fan, Liudmila Zavolokina, Sarah Ebling

The authors test how text-to-image models interpret an underspecified request for “a person with a disability,” comparing generic-prompt images with images generated from named categories through CLIP similarity and human sentiment judgments. Both SDXL and DALL·E 3 matched mobility-impairment images most closely, while raters judged DALL·E 3’s mental-disorder images more negative than SDXL’s; mitigation is not established as the cause.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.