VQA-BC: ROBUST VISUAL QUESTION ANSWERING VIA BIDIRECTIONAL CHAINING

Current VQA models are suffering from the problem of over-dependence on language bias, which severely reduces their robustness in real-world scenarios. In this paper, we analyze VQA models from the view of forward/backward chaining in the inference engine, and propose to enhance their robustness via a novel Bidirectional Chaining (VQA-BC) framework. Specifically, we introduce a backward chaining with hard-negative contrastive learning to reason from the consequence (answers) to generate crucial known facts (question-related visual region features). Furthermore, to alleviate the over-confident problem in answer prediction (forward chaining), we present a novel introspective regularization to connect forward and backward chaining with label smoothing. Extensive experiments verify that VQA-BC not only effectively overcomes language bias on out-of-distribution dataset, but also alleviates the over-correct problem caused by ensemble-based method on in-distribution dataset. Compared with competitive debiasing strategies, our method achieves state-of-the-art performance to reduce language bias on VQA-CP v2 dataset.