papersSEP 10 04:00 UTC
ViSR-KGC: Vision-Language Model Approach to Multimodal Knowledge Graph Completion
Researchers have introduced ViSR-KGC, a method that uses vision-language models to reason over visual subgraphs when filling in missing entities or relations in knowledge graphs. The approach extends knowledge graph completion to multimodal settings by combining textual graph structure with information from entity-associated images. The work is detailed in a revised preprint posted on arXiv.