SciAGENTS: Automating Scientific Discovery Through Multi-Agent Intelligent Graph Reasoning

This work introduces SciAgents, a multi-agent intelligent system leveraging large-scale ontological knowledge graphs, large language models (LLMs), and in-situ learning to automate scientific discovery. Applied to biologically inspired materials, SciAgents reveals hidden interdisciplinary relationships, generating and refining research hypotheses and elucidating underlying mechanisms. The modular framework facilitates material discoveries, hypothesis critique, data retrieval, and identification of research limitations. Case studies demonstrate the system's scalability in combining generative AI, ontological representations, and multi-agent modeling, accelerating materials development by unlocking nature's design principles.

Alireza Ghafarollahi, Markus J. Buehler