Search papers, labs, and topics across Lattice.
This paper introduces DuPLeR, a Dual-Path LLM Reasoning framework designed to enhance multimodal few-shot knowledge graph completion (KGC) by integrating multimodal LLM-derived type priors with factual support structures. The method addresses the challenges of inductive KGC, particularly in scenarios with emerging entities and relations, by employing dual-level structural reasoning and a multimodal enhancement module that regulates message passing. Experimental results on eight inductive variants of two multimodal KG benchmarks demonstrate that DuPLeR significantly improves performance in data-scarce environments, showcasing its robustness and effectiveness.
Robust multimodal few-shot knowledge graph completion is achievable even in data-scarce scenarios with the novel DuPLeR framework.
Knowledge graph completion (KGC) aims to infer missing facts in knowledge graphs (KGs), thereby improving their completeness and supporting downstream intelligent applications. However, emerging entities and relations in real-world deployments make inductive KGC difficult, especially under few-shot and zero-shot settings. Multimodal information and Large Language Model (LLM)-derived priors can enrich sparse relational contexts, but they may also introduce noisy or hallucinated evidence. To address these issues, we propose DuPLeR, a \textbf{Du}al-\textbf{P}ath \textbf{L}LM \textbf{R}easoning framework for multimodal few-shot KGC. DuPLeR builds a calibrated relation graph by combining multimodal LLM-derived type priors with factual support structures, and performs dual-level structural reasoning over the refined relation topology. Moreover, a dual-pathway multimodal enhancement module regulates message passing with query-relevant multimodal signals and supplements entity representations after graph propagation. Experiments on eight inductive variants of two multimodal KG (MMKG) benchmarks show that DuPLeR achieves robust performance in data-scarce KGC scenarios.
