Agentic Exploration of Physics Models

2025-09-29

AI Summary

This work addresses the challenge of “task-agnostic, open-ended automatic discovery of physical laws” in scientific discovery. We propose SciExplorer, a closed-loop scientific exploration agent powered by large language models (LLMs), which requires no domain-specific priors, model fine-tuning, or task-specialized prompting. Instead, it autonomously conducts iterative hypothesis generation, experimental design, and validation in unknown physical systems by leveraging LLM-driven code interpretation for simulation, data analysis, and symbolic regression. To our knowledge, this is the first unified framework spanning diverse domains—including classical mechanical systems, wave equations, and quantum many-body problems. Experiments demonstrate accurate recovery of governing dynamical equations and Hamiltonians, strong generalization across unseen systems, and robustness to noise. Critically, SciExplorer achieves genuinely universal autonomous scientific discovery using only a minimal set of computational tools.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

The process of scientific discovery relies on an interplay of observations, analysis, and hypothesis generation. Machine learning is increasingly being adopted to address individual aspects of this process. However, it remains an open challenge to fully automate the open-ended, heuristic, iterative loop required to discover the laws of an unknown system by exploring it through experiments and analysis, without tailoring the approach to the specifics of a given task. Here, we introduce SciExplorer, an agent that leverages large language model tool-use capabilities to enable free-form exploration of systems without any domain-specific blueprints, and apply it to the exploration of physical systems that are initially unknown to the agent. We test SciExplorer on a broad set of models spanning mechanical dynamical systems, wave evolution, and quantum many-body physics. Despite using a minimal set of tools, primarily based on code execution, we observe impressive performance on tasks such as recovering equations of motion from observed dynamics and inferring Hamiltonians from expectation values. The demonstrated effectiveness of this setup opens the door towards similar scientific exploration in other domains, without the need for finetuning or task-specific instructions.

Problem

• Automating open-ended scientific discovery of unknown physical systems
• Enabling free-form exploration without domain-specific blueprints
• Recovering equations and Hamiltonians from observed system dynamics

Innovation

• Leverages large language model tool-use capabilities
• Enables free-form exploration without domain-specific blueprints
• Uses minimal tools primarily based on code execution