PhysAgent: A Multi-Agent Large Language Model Framework for Physics Data Analysis and Simulation

The inference of physical laws has traditionally relied on human intuition, analytical reasoning, and experimental observation. However, modern physics research is increasingly constrained by challenges such as high specialization, fragmented workflows, and limited computational resources, which impede scientific progress. To address these issues, we introduce PhysAgent , a novel multi-agent system powered by large language models (LLMs), designed to execute end-to-end scientific workflows—from hypothesis generation and computational modeling to data analysis and reporting. PhysAgent’s innovation lies in its specialized agent collaboration: a Mentor Agent guides scientific reasoning through Socratic questioning, a Student Agent handles technical execution (e.g., code implementation, DFT calculations), and a Leader Agent dynamically optimizes task scheduling and resource allocation. Integrating domain-specific tools like first-principles simulations (e.g., Quantum ESPRESSO, VASP), PhysAgent aims to improve reproducibility while maintaining human-in-the-loop refinement. We demonstrate PhysAgent as a proof-of-concept framework for recovering known physical relations from controlled data, including Kepler-law recovery from orbital trajectories and Newton-law recovery from force-motion data. Furthermore, it extends to ab initio materials modeling, automating electronic structure workflows (e.g., GaAs band-structure and density-of-states calculations). In addition, PhysAgent simulates complex real-world phenomena, such as raindrop flow on train windows. By harmonizing LLM-driven planning with domain-specific tools, PhysAgent provides a reproducible multi-agent workflow for physics data analysis and simulation. We provide a demonstration video of PhysAgent to illustrate its workflow and capabilities: https://b23.tv/emOilkC.