AI RESEARCH
SmoGVLM: A Small, Graph-enhanced Vision-Language Model
arXiv CS.CL
•
ArXi:2604.16517v1 Announce Type: cross Large vision-language models (VLMs) achieve strong performance on multimodal tasks but often suffer from hallucination and poor grounding in knowledge-intensive reasoning. We propose SmoGVLM, a small, graph-enhanced VLM that integrates structured knowledge with visual and textual modalities, using Graph Neural Networks. We investigate the effects of our method across a range of model sizes, from tiny (1.3B) to large (13B) models.