This research explores how large language models can effectively process and reason with graph-structured data, opening new possibilities for complex applications.

• Investigates parametric representations of graphs for LLM understanding
• Enables LLMs to work beyond text to handle relational data structures
• Addresses hallucination reduction by incorporating structured knowledge
• Creates pathways for applications in drug discovery and protein design

For biology and medicine, this research could transform drug discovery processes by enabling more sophisticated molecule analysis and protein interaction understanding, potentially accelerating therapeutic development.

What Do LLMs Need to Understand Graphs: A Survey of Parametric Representation of Graphs