arxivcs.IRcs.CV2026-06-27
Multimodal Graph RAG for Long-range Visually Rich Document Understanding
Multimodal large language models (MLLMs) are widely applied to visual document understanding. However, comprehending long documents remains an issue by the limited context window. Though recent multimodal retrieval-augmented generation (MMRAG) can address this challenge by retrie…