Models generate images

In the upper picture, a knowledge update in the model was carried out with the help of ReFACT. On the left, the original images produced by the model. On the right, after editing. The edits also successfully generalize to close formulations, and show that the method succeeds in performing a significant edit in the knowledge encoded in the model. In the bottom picture, the correction of the gender bias when the input is "A developer". Left: before editing with TIME (implicit assumption: A developer is a man). Right: after editing. Courtesy of the Technion

Correcting biases and updating knowledge in models that generate images

  • No comments
"In their training process, models also learn a lot of factual knowledge about the world. For example, models learn the identities of prime ministers, presidents and even actors who played popular characters in TV series. Such models stop
Science website logo
SEARCH