Exploring the Potential of Large Language Models in Graph Generation
Yang Yao Tsinghua University Xin Wang Tsinghua University Corresponding authors. Zeyang Zhang Tsinghua University Yijian Qin Tsinghua University Ziwei Zhang Tsinghua University Xu Chu Tsinghua University Yuekui Yang Tsinghua University Wenwu Zhu Tsinghua University Corresponding authors. Hong Mei Peking University
Abstract
Large language models (LLMs) have achieved great success in many fields, and recent works have studied exploring LLMs for graph discriminative tasks such as node classification. However, the abilities of LLMs for graph generation remain unexplored in the literature. Graph generation requires the LLM to generate graphs with given properties, which has valuable real-world applications such as drug discovery, while tends to be more challenging. In this paper, we propose LLM4GraphGen to explore the ability of LLMs for graph generation with systematical task designs and extensive experiments. Specifically, we propose several tasks tailored with comprehensive experiments to address key questions regarding LLMs’ understanding of different graph structure rules, their ability to capture structural type distributions, and their utilization of domain knowledge for property-based graph generation. Our evaluations demonstrate that LLMs, particularly GPT-4, exhibit preliminary abilities in graph generation tasks, including rule-based and distribution-based generation. We also observe that popular prompting methods, such as few-shot and chain-of-thought prompting, do not consistently enhance per
中文速览
大语言模型（LLM）在文本、代码等领域表现出色，但它们能否直接"画出"满足特定条件的图结构，此前从未被系统研究过。为此，作者提出 LLM4GraphGen 框架，将图生成任务转化为文本提示，设计了三类由易到难的子任务：按规则生成指定结构的图（如树、环、正则图）、从样本中推断图的分布并采样新图、以及利用领域知识生成具有特定性质的分子结构。大量实验表明，GPT-4 在规则型和分布型图生成上具备初步能力，对树和环等简单结构表现较好，但对复杂规则仍力不从心；值得注意的是，少样本提示和思维链等常用技巧并不能稳定提升生成质量，而 LLM 在分子性质导向生成上也展现出一定潜力。这些发现填补了 LLM 图生成能力评估的空白，为后续将大模型引入药物发现等需要图生成的实际场景提供了重要的基准和方向。
原文 arXiv:2403.14358;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2403.14358v1