IQ-VQA: Intelligent Visual Question Answering

Despite tremendous progress in the field of Visual Question Answering, models today still tend to be inconsistent and brittle. Thus, we propose a model-independent cyclic framework which increases consistency robustness any VQA architecture. We train our answer original question, generate an implication based on then learn generated correctly. As part framework, novel generator generates implied questions from question-answer pair. baseline for future works consistency, provide new human-annotated VQA-Implications dataset. The dataset consists 30k implications 3 types - Logical Equivalence, Necessary Condition Mutual Exclusion made validation show that improves by Open image window rule-based dataset, , without degrading their performance.