V-REX: Benchmarking Exploratory Visual Reasoning via Chain-of-Questions

This article introduces V-REX, a benchmark for evaluating visual reasoning capabilities of AI models. The use of "Chain-of-Questions" suggests an approach that breaks down complex visual understanding tasks into a series of simpler, interconnected questions. This method likely aims to assess the model's ability to reason step-by-step and explain its decision-making process. The source being ArXiv indicates this is likely a research paper.