Risk-Tiered Auditing Framework for Gender Bias in T2I Models

A new research paper from arXiv proposes a risk-aligned auditing framework for gender bias in text-to-image (T2I) generative models. The authors argue that existing evaluations are fragmented, with metrics often reported without a shared understanding of what they measure or how results should be interpreted across different deployment contexts. This limits the usefulness of bias measurement for technical auditing and governance. The framework connects risk categories to evaluation metrics, aiming to provide a more structured approach to auditing gender bias in T2I outputs, which are increasingly used in education, media, and public-facing communication.