KnowledgeNavigator: leveraging large language models for enhanced reasoning over knowledge graph

• DOI: 10.1007/s40747-024-01527-8
• Publication Date: 2024-07-02T15:21:21Z
• Authors: Tiezheng Guo; Qingwen Yang; Chen Wang; Yanyi Liu; Pan Li; Jiawei Tang; Dapeng Li; Yingyou Wen
• Keywords: Knowledge graph; Language Understanding; Natural language understanding

Abstract Large language models have achieved outstanding performance on various downstream tasks with their advanced understanding of natural and zero-shot capability. However, they struggle knowledge constraints, particularly in requiring complex reasoning or extended logical sequences. These limitations can affect question answering by leading to inaccuracies hallucinations. This paper proposes a novel framework called KnowledgeNavigator that leverages large graphs achieve accurate interpretable multi-hop reasoning. Especially an analysis-retrieval-reasoning process, searches the optimal path iteratively retrieve external guide reliable answers. treats as flexible components be switched between different without additional costs. Experiments three benchmarks demonstrate significantly improves outperforms all models-based baselines.