ACM SIGKDD 2024
A Survey of Large Language Models for Graphs
Large Language Models for Graphs: Progresses and Directions

1The University of Hong Kong 2Baidu Inc 3University of Notre Dame

Sunday, August 25, 2:00 PM – 5:00 PM | Room 116, KDD 2024 | Barcelona, Spain

About this survey & tutorial

Graphs are an essential data structure utilized to represent relationships in real-world scenarios. Prior research has established that Graph Neural Networks (GNNs) deliver impressive outcomes in graph-centric tasks, such as link prediction and node classification. Despite these advancements, challenges like data sparsity and limited generalization capabilities continue to persist. Recently, Large Language Models (LLMs) have gained attention in natural language processing. They excel in language comprehension and summarization. Integrating LLMs with graph learning techniques has attracted interest as a way to enhance performance in graph learning tasks.

We will explore four primary categories of methods that harness the power of large language models (LLMs) in graph tasks. These are: i) Graph Neural Networks (GNNs) as Prefix, ii) LLMs as Prefix, iii) LLMs-Graphs Integration, and iv) LLMs-Only. In each section, we will introduce you to the leading techniques in this field and provide you with sample code snippets to experiment with. This tutorial is designed to be valuable for both researchers aiming to pioneer new LLM4Graph solutions and industry professionals seeking to apply these methods in practical, real-world scenarios.

Schedule [Full Slides] [Open Resources on Github]

Time Section Presenter
14:00 - 14:10 Opening & Introduction Prof. Chao Huang
14:10 - 14:40 Section 1: GNNs as Prefix Jiabin Tang
14:40 - 14:45 Q & A for Section 1 Jiabin Tang
14:45 - 15:15 Section 2: LLMs as Prefix Lianghao Xia
15:15 - 15:20 Q & A for Section 2 Lianghao Xia
15:20 - 15:50 Section 3: LLMs-Graphs Intergration Xubin Ren
15:50 - 15:55 Q & A for Section 3 Xubin Ren
15:55 - 16:25 Coffee Break -
16:25 - 16:55 Section 4: LLMs-Only Jiabin Tang & Xubin Ren
16:55 - 17:00 Q & A for Section 4 Jiabin Tang & Xubin Ren

Reading List

Works in our group are highlighted in bold.


Section 1: GNNs as Prefix

3.1 Node- level Tokenization


3.1 Node- level Tokenization


Section 2: LLMs as Prefix

2.1 Embeddings from LLMs for GNNs


2.2 Labels from LLMs for GNNs


Section 3: LLMs-Graphs Intergration

3.1 Alignment between GNNs and LLMs


3.2 Fusion Training of GNNs and LLMs


3.3 LLMs Agent for Graphs


Section 4: LLMs-Only

4.1 Tuning-free


4.2 Tuning-required


BibTeX

@inproceedings{ren2024survey,
  title={A survey of large language models for graphs},
  author={Ren, Xubin and Tang, Jiabin and Yin, Dawei and Chawla, Nitesh and Huang, Chao},
  booktitle={Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={6616--6626},
  year={2024}
}

@inproceedings{huang2024large,
  title={Large Language Models for Graphs: Progresses and Directions},
  author={Huang, Chao and Ren, Xubin and Tang, Jiabin and Yin, Dawei and Chawla, Nitesh},
  booktitle={Companion Proceedings of the ACM on Web Conference 2024},
  pages={1284--1287},
  year={2024}
}