GPT-5 Agent Autonomously Derives Classical and Quantum Physics Models

jwt0625 · x · 2026-08-06

A recent study titled 'Agentic Exploration of Physics Models' demonstrates the potential of LLMs in scientific research workflows. The author notes that by using a simple prompt to guide an agent through typical scientific workflows, LLMs are getting quite good at backing out the physics model from observations for classical mechanical systems and simple quantum systems.

However, the models remain less consistent when dealing with quantum many-body physics and data that includes noise.

Related event: GPT-5 Excels at Classical Physics but Struggles with Quantum Many-Body Problems(2 posts)→

Original post →

More from coding & agent

coding & agent channel →