Ask4VG: Risk-Aware Question Selection for Reducing Prior-Driven Answers in Medical VQA 事件

Ask4VG: Risk-Aware Question Selection for Reducing Prior-Driven Answers in Medical VQA arXiv:2606.01044v1 Announce Type: new Abstract: Medical visual question answering requires models to ground their responses in image evidence, because visually unsupported answers can mislead downstream interpretation. However, many medical VQA questions are generic, template-like, or highly similar in form, which can encourage models to learn question-answer shortcuts instead of image-dependent reasoning and