With the rise of large language models (LLMs), there has been growing interest in Graph Foundation Models (GFMs) for graph-based tasks. By leveraging LLMs as predictors, GFMs have demonstrated impressive generalizability across various tasks and datasets. However, existing research on LLMs as predictors has predominantly focused on static graphs, leaving their potential in dynamic graph prediction unexplored. In this work, we pioneer using LLMs for predictive tasks on dynamic graphs. We identify two key challenges: the constraints imposed by context length when processing large-scale historical data and the significant variability in domain characteristics, both of which complicate the development of a unified predictor. To address these challenges, we propose the GraphAgent-Dynamic (GAD) Framework, a multi-agent system that leverages collaborative LLMs. In contrast to using a single LLM as the predictor, GAD incorporates global and local summary agents to generate domain-specific knowledge, enhancing its transferability across domains. Additionally, knowledge reflection agents enable adaptive updates to GAD's knowledge, maintaining a unified and self-consistent architecture. In experiments, GAD demonstrates performance comparable to or even exceeds that of full-supervised graph neural networks without dataset-specific training. Finally, to enhance the task-specific performance of LLM-based predictors, we discuss potential improvements, such as dataset-specific fine-tuning to LLMs. By developing tailored strategies for different tasks, we provide new insights for the future design of LLM-based predictors.
Graph-structured data underpin a wide range of real-world systems, from molecular chemistry and biological interactions to social and information networks. Recent advancements in deep generative models, along with the emergence of large language models (LLMs), have spurred significant progress in graph generative models (GGMs), enabling the synthesis of complex and realistic graph structures. This article provides a comprehensive overview of the literature in this field. We begin by distinguishing between two primary categories of graph datasets based on their structural formation mechanisms: geometric and scale-free. Building on this foundation, we propose a unified taxonomy that systematically organizes this rapidly evolving landscape by jointly considering (i) the category of generated graphs, (ii) the graph attribute modality, and (iii) the underlying probabilistic modeling paradigm. We then analyze representative neural network architectures and modeling strategies, followed by an overview of evaluation metrics. Finally, we highlight key applications in molecular design, protein optimization, social network analysis, and recommendation systems and outline four promising directions for future research.