Do Vision-Language Models Really Understand Visual Language?
This research addresses the critical problem of evaluating the true visual language understanding of LVLMs, which is important for researchers and practitioners developing and deploying these models in domains requiring diagram interpretation.
This paper investigates the diagram comprehension capabilities of Large Vision-Language Models (LVLMs) using a comprehensive test suite. The study found that while LVLMs can accurately identify and reason about entities, their understanding of relationships within diagrams is limited, with impressive performance often stemming from leveraging background knowledge rather than genuine visual language comprehension.
Visual language is a system of communication that conveys information through symbols, shapes, and spatial arrangements. Diagrams are a typical example of a visual language depicting complex concepts and their relationships in the form of an image. The symbolic nature of diagrams presents significant challenges for building models capable of understanding them. Recent studies suggest that Large Vision-Language Models (LVLMs) can even tackle complex reasoning tasks involving diagrams. In this paper, we investigate this phenomenon by developing a comprehensive test suite to evaluate the diagram comprehension capability of LVLMs. Our test suite uses a variety of questions focused on concept entities and their relationships over a set of synthetic as well as real diagrams across domains to evaluate the recognition and reasoning abilities of models. Our evaluation of LVLMs shows that while they can accurately identify and reason about entities, their ability to understand relationships is notably limited. Further testing reveals that the decent performance on diagram understanding largely stems from leveraging their background knowledge as shortcuts to identify and reason about the relational information. Thus, we conclude that LVLMs have a limited capability for genuine diagram understanding, and their impressive performance in diagram reasoning is an illusion emanating from other confounding factors, such as the background knowledge in the models.