AI RESEARCH
Ask4VG: Risk-Aware Question Selection for Reducing Prior-Driven Answers in Medical VQA
arXiv CS.CV
•
ArXi:2606.01044v1 Announce Type: new Medical visual question answering requires models to ground their responses in image evidence, because visually uned 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 thereby increase the risk of hallucinated responses. We propose Ask4VG, a label-free pilot framework for risk-aware question selection.