# Exploring Large Language Models for Knowledge Graph Completion

**Authors:** Liang Yao, Jiazhen Peng, Chengsheng Mao, Yuan Luo  
**arXiv:** [2308.13916](https://arxiv.org/abs/2308.13916)  
**Date:** 2023-08-26  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.13916v4/

## Key points

- Knowledge graphs are crucial for various artificial intelligence tasks
- Common challenge of incompleteness in knowledge graphs addressed
- Proposal of using Large Language Models (LLM) for knowledge graph completion
- Introduction of Knowledge Graph LLM (KG-LLM) framework to model triples effectively
- Leveraging entity and relation descriptions for predictions
- Achieving state-of-the-art performance in triple classification and relation prediction tasks
- Fine-tuning smaller models like LLaMA-7B and ChatGLM-6B surpasses larger models such as ChatGPT and GPT-4

## Summary

In their study titled "Exploring Large Language Models for Knowledge Graph Completion," authors Liang Yao, Jiazhen Peng, Chengsheng Mao, and Yuan Luo delve into the crucial role that knowledge graphs play in various artificial intelligence tasks. They address the common challenge of incompleteness faced by these knowledge graphs and propose a novel approach using Large Language Models (LLM) for knowledge graph completion. The researchers treat triples within knowledge graphs as text sequences and introduce a cutting-edge framework known as Knowledge Graph LLM (KG-LLM) to effectively model these triples. By leveraging entity and relation descriptions of a triple as prompts, the technique generates responses used for predictions. Through extensive experiments conducted on diverse benchmark knowledge graphs, the team demonstrates that their method achieves state-of-the-art performance in tasks such as triple classification and relation prediction. Notably, they find that fine-tuning relatively smaller models like LLaMA-7B and ChatGLM-6B surpasses the performance of recent models such as ChatGPT and GPT-4. This research represents a significant advancement in utilizing Large Language Models for enhancing the completeness of knowledge graphs, showcasing promising results in improving AI applications reliant on structured data representations.