NODES AI 2026 - Agent Reasoning with Graph World Models

In this talk, Lucas Godfrey explores the concept of agent reasoning with graph world models, focusing on their application in robotics and spatial scenarios. The session emphasizes the importance of graph representations in world model design, and discusses how they can enhance the performance of large language model-based agents. Attendees will learn about modeling states, actions, and environments, as well as strategies for iterating on existing graph schemas and agent architectures. This presentation is beneficial for both developers actively implementing these concepts and those looking to understand how world models can shape their initial designs for agent systems.