Reconstruction as a Bridge for Event-Based Visual Question Answering

This article likely discusses a novel approach to visual question answering (VQA) that leverages reconstruction techniques. The focus is on event-based VQA, suggesting the system is designed to understand and answer questions about events depicted in visual data. The use of 'reconstruction' implies the system might attempt to reconstruct the visual scene or event to better understand it and answer questions. The ArXiv source indicates this is a research paper.