Improving Complex Reasoning over Knowledge Graph with Logic-Aware Curriculum Tuning

TL;DR

This paper introduces LACT, a curriculum-based instruction tuning framework for large language models that enhances complex reasoning over knowledge graphs, significantly improving performance on logical query answering tasks.

Contribution

It proposes a novel logic-aware curriculum tuning method that leverages binary tree decomposition of queries to boost LLM reasoning over incomplete knowledge graphs.

Findings

LACT achieves an average +5.5% MRR improvement.

It sets new state-of-the-art results on multiple datasets.

The framework effectively addresses reasoning difficulty gaps.

Abstract

Answering complex queries over incomplete knowledge graphs (KGs) is a challenging job. Most previous works have focused on learning entity/relation embeddings and simulating first-order logic operators with various neural networks. However, they are bottlenecked by the inability to share world knowledge to improve logical reasoning, thus resulting in suboptimal performance. In this paper, we propose a complex reasoning schema over KG upon large language models (LLMs), containing a curriculum-based logical-aware instruction tuning framework, named LACT. Specifically, we augment the arbitrary first-order logical queries via binary tree decomposition, to stimulate the reasoning capability of LLMs. To address the difficulty gap among different types of complex queries, we design a simple and flexible logic-aware curriculum learning framework. Experiments across widely used datasets demonstrate that LACT has substantial improvements~(brings an average +5.5% MRR score) over advanced methods, achieving the new state-of-the-art.