KG-CF: Knowledge Graph Completion with Context Filtering under the Guidance of Large Language Models

2025-01-06

AI Summary

Existing knowledge graph completion (KGC) methods predominantly employ large language models (LLMs) for classification rather than entity ranking, limiting their ability to produce high-confidence predictions; while incorporating graph paths improves accuracy, it often introduces noise. This paper proposes KG-CF, the first framework to deeply integrate LLMs into ranking-based KGC. It introduces a context filtering mechanism that dynamically selects highly relevant graph paths via prompt engineering and path importance scoring, effectively suppressing redundant and noisy signals. Furthermore, it establishes an end-to-end differentiable architecture for path-aware candidate entity reranking. Evaluated on FB15k-237 and WN18RR, KG-CF achieves new state-of-the-art performance, improving mean reciprocal rank (MRR) and Hits@1 by 4.2% and 5.8% on average, respectively—overcoming the prevailing limitation of using LLMs solely as classifiers in KGC.

Technology Category

• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Machine Learning: Learning Preferences or Rankings
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Large Language Models (LLMs) have shown impressive performance in various tasks, including knowledge graph completion (KGC). However, current studies mostly apply LLMs to classification tasks, like identifying missing triplets, rather than ranking-based tasks, where the model ranks candidate entities based on plausibility. This focus limits the practical use of LLMs in KGC, as real-world applications prioritize highly plausible triplets. Additionally, while graph paths can help infer the existence of missing triplets and improve completion accuracy, they often contain redundant information. To address these issues, we propose KG-CF, a framework tailored for ranking-based KGC tasks. KG-CF leverages LLMs' reasoning abilities to filter out irrelevant contexts, achieving superior results on real-world datasets. The code and datasets are available at url{https://anonymous.4open.science/r/KG-CF}.

Problem

• Knowledge Graph Completion
• Entity Ranking
• Information Noise

Innovation

• KG-CF framework
• Large Language Models
• Ranking Task