William A. Ingram

VisURLX is presented, a pipeline built on a multimodal large language model (MLLM) that reads each rendered page image directly, and extracts each URL with its target sentence and surrounding context across the body, footnotes, and references.