A structure-anchored multi-modal knowledge graph completion framework that unifies visual and textual tokens into a shared space and aligns both modalities toward structural …
A noise-enhanced multi-modal knowledge graph completion framework that combines adaptive noise sampling with an attention penalty strategy to mitigate attention over-trust and …
A large-model-driven framework for balanced multi-modal knowledge graph completion, combining SDXL-based visual completion, CLIP contrastive cross-modal alignment, and …