Counterfactual Visual Dialog: Robust Commonsense Knowledge Learning From Unbiased Training | IEEE Journals & Magazine | IEEE Xplore