1 paper
Anne Harrington, A. Sophia Koepke, Shyamgopal Karthik +2
Contemporary text-to-image models exhibit a surprising degree of mode collapse, as can be seen when sampling several images given the same text prompt. Previous work has attempted…