Is the O3 model trained on the ARC public dataset? And is that the reason it performed well for the ARC Prize?

Comments