Dual-Path LLM Reasoning for Multimodal Few-Shot Knowledge Graph Completion

2026-07-29

AI Summary

This work addresses the challenges of multimodal noise and large language model (LLM) hallucinations in inductive knowledge graph completion under few-shot and zero-shot settings. To this end, the authors propose DuPLeR, a novel framework that calibrates LLM-generated relation priors with factual structures to construct a reliable relation graph and employs a dual-level graph neural network for structural reasoning. Additionally, DuPLeR incorporates a dual-path multimodal augmentation mechanism that integrates query-relevant multimodal signals during graph propagation to refine entity representations. Evaluated across eight inductive settings on two multimodal knowledge graph benchmarks, DuPLeR substantially outperforms existing methods, demonstrating enhanced robustness and accuracy in knowledge graph completion under data-scarce conditions.

Abstract

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.

Problem

• Knowledge Graph Completion
• Few-Shot Learning
• Multimodal Reasoning
• Inductive Inference
• LLM Hallucination

Innovation

• Dual-Path Reasoning
• Multimodal Knowledge Graph Completion
• Few-Shot Learning
• Large Language Models
• Inductive KGC