"We are also releasing three new datasets: Screen Annotation to evaluate the layout understanding capability of the model, as well as ScreenQA Short and Complex

"We are also releasing three new datasets: Screen Annotation to evaluate the layout understanding capability of the model, as well as ScreenQA Short and Complex ScreenQA for a more comprehensive evaluation of its QA capability." Looks useful to me for replicating some things. Good stuff!