Chart Question Answering

The task of answering natural-language questions about a data visualization, typically a chart provided as an image or structured specification. A chart question answering system must identify the chart type, extract the underlying data, interpret axes and legends, and answer questions spanning value retrieval (what is the value of X?), extrema (which category is highest?), comparison, trend, and computation. In accessibility contexts, chart question answering is a promising alternative to static alt-text descriptions or flattened data tables, because it lets blind and low-vision users interactively explore a visualization and ask only the questions they care about. Benchmarks include VLAT, ChartQA, ChartMuseum, and EncQA; recent systems use large vision-language models with prompting strategies such as Charts-of-Thought.