Can LLM Graph Reasoning Generalize beyond Pattern Memorization?

TL;DR

This paper evaluates whether large language models can generalize graph reasoning skills beyond pattern memorization, revealing limitations and exploring strategies to improve their real-world reasoning capabilities.

Contribution

It introduces the NLGift benchmark for assessing LLM graph reasoning generalization and analyzes the effectiveness of different strategies to enhance this ability.

Findings

LLMs perform well on simple pattern generalization

Struggle to generalize across complex reasoning and real-world data

Post-training alignment shows promise for real-world tasks

Abstract

Large language models (LLMs) demonstrate great potential for problems with implicit graphical structures, while recent works seek to enhance the graph reasoning capabilities of LLMs through specialized instruction tuning. The resulting 'graph LLMs' are evaluated with in-distribution settings only, thus it remains underexplored whether LLMs are learning generalizable graph reasoning skills or merely memorizing patterns in the synthetic training data. To this end, we propose the NLGift benchmark, an evaluation suite of LLM graph reasoning generalization: whether LLMs could go beyond semantic, numeric, structural, reasoning patterns in the synthetic training data and improve utility on real-world graph-based tasks. Extensive experiments with two LLMs across four graph reasoning tasks demonstrate that while generalization on simple patterns (semantic, numeric) is somewhat satisfactory, LLMs struggle to generalize across reasoning and real-world patterns, casting doubt on the benefit of synthetic graph tuning for real-world tasks with underlying network structures. We explore three strategies to improve LLM graph reasoning generalization, and we find that while post-training alignment is most promising for real-world tasks, empowering LLM graph reasoning to go beyond pattern memorization remains an open research question.