How well does GPT-4V perform Visual Question Answering (VQA) on Data Structures?
There is no dataset for VQA on graph and tree data structures in previous work, so we must create one. We create a standard, repeatable process for selecting and obtaining VQA tasks that must fall under a certain criteria.
Technical overview of creation and evaluation of dataset.
We elect GPT-4V as the primary model to observe for this evaluation.
Overview of the dataset architecture.
