ColGraphRAG: Late-Interaction Evidence Retrieval for Multimodal GraphRAG

A new arXiv preprint introduces ColGraphRAG, a late-interaction evidence retrieval method for multimodal GraphRAG that replaces single-vector bi-encoder similarity with MaxSim-style multi-vector scoring for graph-linked image nodes. On MultimodalQA, the approach yields improved retrieval-stage point estimates for graph-linked image candidates and downstream QA gains, especially where visual evidence matters most, though mixed trends on text-dominant questions suggest broader validation is needed.

arXiv:2607.16208v1 Announce Type: new Abstract: Graph-grounded multimodal question answering organizes text, tables, and images in a structured evidence graph, yet end-to-end accuracy depends on which multimodal assets are ranked highly enough to enter downstream reasoning; for graph-linked images, single-vector bi-encoder similarity can discard patch- and token-level structure needed for fine-grained alignment. We evaluate replacing the visual candidate-ranking operator over graph-linked image nodes with late-interaction MaxSim-style multi-vector scoring in the ColBERT/ColPali lineage, while keeping offline graph construction, text- and table-side retrieval, structured extraction, and downstream reasoning unchanged. On MultimodalQA, this change is associated with improved retrieval-stage point estimates for graph-linked image candidates and downstream QA gains, with larger movement where visual evidence matters most and mixed trends on text-dominant questions; we interpret the pattern as mechanism-level evidence for graph-linked visual evidence inclusion, while broader validation and finer graph-level diagnostics remain important future work.