CVPR'23

In this half-day CVPR'23 tutorial, we present the state-of-the-art works on attention explainability and probing. We demonstrate how these mechanisms can be leveraged to guide diffusion models to edit and correct their generated images.