【KDD2024教程】《大规模语言模型在图上的综述: 进展与方向》，142页ppt

图是一种用于表示现实世界中关系的关键数据结构。先前的研究已经证明，图神经网络（GNNs）在以图为中心的任务中（如链接预测和节点分类）表现出色。尽管取得了这些进展，但数据稀疏性和有限的泛化能力等挑战依然存在。近年来，大规模语言模型（LLMs）在自然语言处理领域引起了广泛关注。它们在语言理解和总结方面表现优异。将LLMs与图学习技术相结合，作为提升图学习任务性能的一种方法，逐渐引起了研究兴趣。我们将探讨利用大规模语言模型（LLMs）在图任务中发挥作用的四类主要方法。这四类方法分别是：i) 图神经网络（GNNs）作为前缀，ii) LLMs作为前缀，iii) LLMs与图的集成，以及iv) 仅使用LLMs。在每个部分中，我们将向您介绍该领域的领先技术，并提供示例代码片段供您实验。本教程旨在为希望开创新的LLM4Graph解决方案的研究人员，以及希望在实际场景中应用这些方法的行业专业人士提供有价值的参考。参考文献:

Section 1: GNNs as Prefix

**3.1 Node- level Tokenization

GraphGPT: Graph instruction tuning for large language models
HiGPT: Heterogeneous Graph Language Model
GraphTranslator: Aligning Graph Model to Large Language Model for Open-ended Tasks
UniGraph: Learning a Cross-Domain Graph Foundation Model From Natural Language
GIMLET:Aunifiedgraph-textmodelforinstruction-based molecule zero-shot learning
XRec: Large Language Models for Explainable Recommendation

**3.1 Node- level Tokenization

GraphLLM: Boosting graph reasoning ability of large language model
GIT-Mol: A multi-modal large language model for molecular science with graph, image, and text
MolCA: Molecular graph-language modeling with cross-modal projector and uni-modal adapter
InstructMol: Multi-modal integration for building a versatile and reliable molecular assistant in drug discovery
G-Retriever: Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering
Graph neural prompting with large language models

Section 2: LLMs as Prefix

**2.1 Embeddings from LLMs for GNNs

Prompt-based node feature extractor for few-shot learning on text-attributed graphs
SimTeG: A frustratingly simple approach improves textual graph learning
Graph-aware language model pre-training on a large graph corpus can help multiple graph applications
One for all: Towards training one graph model for all classification tasks
Harnessing explanations: Llm-to-lm interpreter for enhanced text-attributed graph representation learning
LLMRec: Large language models with graph augmentation for recommendation

**2.2 Labels from LLMs for GNNs

OpenGraph: Towards Open Graph Foundation Models
Label-free node classification on graphs with large language models (LLMs)
GraphEdit: Large Language Models for Graph Structure Learning
Representation learning with large language models for recommendation

Section 3: LLMs-Graphs Intergration

**3.1 Alignment between GNNs and LLMs

A molecular multimodal foundation model associating molecule graphs with natural language
ConGraT: Self-supervised contrastive pretraining for joint graph and text embeddings
Prompt tuning on graph-augmented low-resource text classification
GRENADE: Graph-Centric Language Model for Self-Supervised Representation Learning on Text-Attributed Graphs
Multi-modal molecule structure–text model for text-based retrieval and editing
Pretraining language models with text-attributed heterogeneous graphs
Learning on large-scale text-attributed graphs via variational inference

**3.2 Fusion Training of GNNs and LLMs

GreaseLM: Graph reasoning enhanced language models for question answering
Disentangled representation learning with large language models for text-attributed graphs
Efficient Tuning and Inference for Large Language Models on Textual Graphs
Can GNN be Good Adapter for LLMs?

**3.3 LLMs Agent for Graphs

Don't Generate, Discriminate: A Proposal for Grounding Language Models to Real-World Environments
Graph Agent: Explicit Reasoning Agent for Graphs
Middleware for LLMs: Tools Are Instrumental for Language Agents in Complex Environments
Call Me When Necessary: LLMs can Efficiently and Faithfully Reason over Structured Environments
Reasoning on graphs: Faithful and interpretable large language model reasoning

Section 4: LLMs-Only

**4.1 Tuning-free

Can language models solve graph problems in natural language?
GPT4Graph: Can large language models understand graph structured data? an empirical evaluation and benchmarking
BeyondText: A Deep Dive into Large Language Models’ Ability on Understanding Graph Data
Exploring the potential of large language models (llms) in learning on graphs
Graphtext: Graph reasoning in text space
Talk like a graph: Encoding graphs for large language models
LLM4DyG: Can Large Language Models Solve Problems on Dynamic Graphs?
Which Modality should I use–Text, Motif, or Image?: Understanding Graphs with Large Language Models
When Graph Data Meets Multimodal: A New Paradigm for Graph Understanding and Reasoning

**4.2 Tuning-required

Natural language is all a graph needs
Walklm: A uniform language model fine-tuning framework for attributed graph embedding
LLaGA: Large Language and Graph Assistant
InstructGraph: Boosting Large Language Models via Graph-centric Instruction Tuning and Preference Alignment
ZeroG: Investigating Cross-dataset Zero-shot Transferability in Graphs
GraphWiz: An Instruction-Following Language Model for Graph Problems
GraphInstruct: Empowering Large Language Models with Graph Understanding and Reasoning Capability
MuseGraph: Graph-oriented Instruction Tuning of Large Language Models for Generic Graph Mining