Integrating Graphs, Large Language Models, and Agents: Reasoning and Retrieval
Summary: arXiv:2604.15951v1 Announce Type: new
Abstract: Generative AI, particularly Large Language Models, increasingly integrates graph-based representations to enhance reasoning, retrieval, and structured decision-making. Despite rapid advances, there remains limited clarity regarding when, why, where, and what types of graph-LLM integrations are most appropriate across applications. This survey provides a concise, structured overview of the design choices underlying the integration of graphs with LLMs.
Key Highlights
- Purpose Categorization: The integration methods are categorized based on their objectives, including:
- Reasoning
- Retrieval
- Generation
- Recommendation
- Graph Modalities: Various types of graph-based representations are discussed, including:
- Knowledge Graphs
- Scene Graphs
- Interaction Graphs
- Causal Graphs
- Dependency Graphs
- Integration Strategies: The survey outlines several strategies for integrating graphs with LLMs, such as:
- Prompting
- Augmentation
- Training
- Agent-based use
Application Domains
The survey maps representative works across various domains, including:
- Cybersecurity: Enhancing threat detection and response capabilities.
- Healthcare: Improving patient diagnosis and treatment recommendations.
- Materials Science: Accelerating the discovery of new materials and their properties.
- Finance: Supporting risk assessment and investment strategies.
- Robotics: Facilitating autonomous decision-making and navigation.
- Multimodal Environments: Integrating diverse data sources for richer insights.
Strengths and Limitations
The survey emphasizes the strengths and limitations of each technique, providing insights into:
- The effectiveness of different graph-LLM integrations based on task requirements.
- The suitability of various graph modalities for specific applications.
- The complexity of reasoning tasks and how it affects integration choices.
Conclusion
This survey aims to offer researchers a practical guide for selecting the most suitable graph-LLM approach depending on task requirements, data characteristics, and reasoning complexity. By clarifying the landscape of graph and LLM integration, it serves as a valuable resource for advancing the field of generative AI.