Learning to Plan for Retrieval-Augmented Large Language Models from Knowledge Graphs

TL;DR

This paper presents a novel framework that leverages knowledge graphs to generate planning data, enhancing the ability of fine-tuned large language models to decompose and answer complex questions involving retrieval.

Contribution

The paper introduces a new method for generating planning data from knowledge graphs to improve LLMs' planning and retrieval capabilities, reducing reliance on manual annotation.

Findings

Enhanced LLM planning abilities with KG-derived data

Improved performance on complex QA tasks

Effective across multiple datasets and a new benchmark

Abstract

Improving the performance of large language models (LLMs) in complex question-answering (QA) scenarios has always been a research focal point. Recent studies have attempted to enhance LLMs' performance by combining step-wise planning with external retrieval. While effective for advanced models like GPT-3.5, smaller LLMs face challenges in decomposing complex questions, necessitating supervised fine-tuning. Previous work has relied on manual annotation and knowledge distillation from teacher LLMs, which are time-consuming and not accurate enough. In this paper, we introduce a novel framework for enhancing LLMs' planning capabilities by using planning data derived from knowledge graphs (KGs). LLMs fine-tuned with this data have improved planning capabilities, better equipping them to handle complex QA tasks that involve retrieval. Evaluations on multiple datasets, including our newly proposed benchmark, highlight the effectiveness of our framework and the benefits of KG-derived planning data.