Visual Entailment

AbstractWhile sophisticated neural-based models have achieved remarkable success in Visual Question Answering (VQA), these models tend to answer questions only according to superficial correlations between question and answer. Several recent …...