Enhancing LLM Tool Use with High-quality Instruction Data from Knowledge Graph

TL;DR

This paper introduces a novel approach using knowledge graphs to generate high-quality instruction data, significantly enhancing large language models' ability to utilize tools effectively and improve problem-solving skills.

Contribution

It presents a new method leveraging knowledge graphs to produce superior instruction data for fine-tuning LLMs, outperforming previous data generation techniques.

Findings

Improved tool utilization in LLMs after fine-tuning

Synthetic data from knowledge graphs enhances problem-solving

Small data samples yield significant performance gains

Abstract

Teaching large language models (LLMs) to use tools is crucial for improving their problem-solving abilities and expanding their applications. However, effectively using tools is challenging because it requires a deep understanding of tool functionalities and user intentions. Previous methods relied mainly on LLMs to generate instruction data, but the quality of these data was often insufficient. In this paper, we propose a new method that uses knowledge graphs to generate high-quality instruction data for LLMs. Knowledge graphs are manually curated datasets rich in semantic information. We begin by extracting various query pathways from a given knowledge graph, which are transformed into a broad spectrum of user queries. We then translate the relationships between entities into actionable tools and parse the pathways of each query into detailed solution steps, thereby creating high-quality instruction data. Our experiments show that fine-tuning on just a small sample of this synthetic data can significantly improve the tool utilization and overall capabilities of LLMs.