Visual Grounding Methods for VQA are Working for the Wrong Reasons!

arXiv:2004.05704v3 Announce Type: replace-cross
Abstract: Existing Visual Question Answering (VQA) methods tend to exploit dataset biases and spurious statistical correlations, instead of producing right answers for the right reasons. To address this issue, recent bias mitigation methods for VQA propose to incorporate visual cues (e.g., human attention maps) to better ground the VQA models, showcasing impressive gains. However, we show that the performance improvements are not a result of improved visual grounding, but a regularization effect which prevents over-fitting to …

abstract arxiv attention bias biases correlations cs.ai cs.cl cs.cv dataset exploit human issue maps question question answering statistical type visual visual cues vqa