A Survey of Graph Meets Large Language Model: Progress and Future Directions

TL;DR

This survey systematically categorizes how large language models can augment graph learning, proposing a taxonomy that classifies methods into LLMs as enhancers, predictors, and alignment components. It reviews explanation- and embedding-based enhancement, flattening- and GNN-based prediction, and symmetric/asymmetric alignment strategies, with representative examples and notable caveats such as data leakage and efficiency. The work identifies key limitations and articulates future directions, including handling non-text-attributed graphs, improving transferability, and developing graph-aware LLMs or agent-based systems. By consolidating diverse approaches and benchmarking considerations, the paper clarifies how text-centric models can enrich graph representations and reasoning, while highlighting computational and methodological challenges to be addressed. Overall, the survey provides a structured roadmap for advancing graph learning with LLMs and informs researchers and practitioners about promising directions and practical trade-offs.

Abstract

Graph plays a significant role in representing and analyzing complex relationships in real-world applications such as citation networks, social networks, and biological data. Recently, Large Language Models (LLMs), which have achieved tremendous success in various domains, have also been leveraged in graph-related tasks to surpass traditional Graph Neural Networks (GNNs) based methods and yield state-of-the-art performance. In this survey, we first present a comprehensive review and analysis of existing methods that integrate LLMs with graphs. First of all, we propose a new taxonomy, which organizes existing methods into three categories based on the role (i.e., enhancer, predictor, and alignment component) played by LLMs in graph-related tasks. Then we systematically survey the representative methods along the three categories of the taxonomy. Finally, we discuss the remaining limitations of existing studies and highlight promising avenues for future research. The relevant papers are summarized and will be consistently updated at: https://github.com/yhLeeee/Awesome-LLMs-in-Graph-tasks.

Figures (5)

• Figure 1: Across a myriad of graph domains, the integration of graphs and LLMs demonstrates success in various downstream tasks.
• Figure 2: A taxonomy of models for solving graph tasks with the help of large language models (LLMs) with representative examples.
• Figure 3: The illustration of LLM-as-enhancer approaches: a) explanation-based enhancement, which uses LLMs to generate explanations of text attributes to enhance text embeddings; b) Embedding-based enhancement, which directly obtains text embeddings by LLMs as initial node embeddings.
• Figure 4: The illustration of LLM-as-predictor approaches: a) Flatten-based prediction, which incorporates graph structure with LLMs via different flattening strategies; b) GNN-based prediction, utilizing GNNs to capture structural information for LLMs.
• Figure 5: The illustration of GNN-LLM-Alignment approaches: a) Contrastive, symmetric alignment which applies concatenation or contrastive learning to graph embeddings and text embeddings; b) Iterative, belongs to symmetric alignment, aiming to implement iterative interactions on embeddings of two modalities; c) Graph-nested, a symmetric alignment which interweaves GNNs with Transformers and d) Distillation, belongs to asymmetric alignment, which uses GNN as a teacher to train language models to be graph-aware.