DuPLeR Framework Improves Multimodal Few-Shot Knowledge Graph Completion

DuPLeR employs dual-pathway reasoning to combine multimodal LLM type priors with factual structural support for knowledge graph completion. The method uses a calibrated relation graph and a multimodal enhancement module to mitigate hallucinations and noise in sparse relational contexts.