VisionTrap: Unanswerable Questions On Visual Data
This addresses the issue of overconfident AI responses in VQA for users relying on accurate visual information, though it is incremental as it focuses on a specific dataset and evaluation aspect.
The research tackled the problem of how Visual Question Answering (VQA) models handle unanswerable questions on visual data, finding that models often fail to recognize their limitations and generate incorrect answers instead of abstaining.
Visual Question Answering (VQA) has been a widely studied topic, with extensive research focusing on how VLMs respond to answerable questions based on real-world images. However, there has been limited exploration of how these models handle unanswerable questions, particularly in cases where they should abstain from providing a response. This research investigates VQA performance on unrealistically generated images or asking unanswerable questions, assessing whether models recognize the limitations of their knowledge or attempt to generate incorrect answers. We introduced a dataset, VisionTrap, comprising three categories of unanswerable questions across diverse image types: (1) hybrid entities that fuse objects and animals, (2) objects depicted in unconventional or impossible scenarios, and (3) fictional or non-existent figures. The questions posed are logically structured yet inherently unanswerable, testing whether models can correctly recognize their limitations. Our findings highlight the importance of incorporating such questions into VQA benchmarks to evaluate whether models tend to answer, even when they should abstain.