pipette
ESEspañol

Experimental reasoning in large language models

J. Kim, P. A. Romero

Preprint

In the authors' words

Scientific reasoning requires interaction with the physical world: designing experiments, interpreting outcomes, and deciding what experiment should come next. Here we show that general-purpose large language models (LLMs) already possess substantial capabilities for experimental reasoning. We connected LLMs to a liquid-handling robot and camera through PEARL, which provides perception, action, and persistent state. Given objectives rather than prescribed workflows, the system designed and successfully executed experiments, interpreted observations, and revised its approach as evidence accumulated. When information was insufficient, PEARL autonomously designed and performed intermediate experiments to acquire it. In quantitative enzyme characterization, it judged whether measurements were sufficient, new data were required, or conditions needed to change. Across eight scenarios, these behaviors were reliable with limited agent scaffolding and no supplied scenario-specific recovery recipes. More broadly, general-purpose LLMs may provide a reasoning layer between scientific objectives and physical experimentation, shifting laboratory automation from executing predefined procedures toward AI systems that participate directly in the experimental process.

Main resultThe abstract does not state a limitation.

In plain words AI summary

The authors connected large language models to a liquid-handling robot and camera through a system called PEARL. Given objectives rather than step-by-step instructions, the system designed and ran experiments, interpreted results, and adjusted its approach based on new evidence. When information was lacking, PEARL independently designed and performed additional experiments to obtain it. In tests of enzyme properties, the system decided whether measurements were enough, more data were needed, or conditions should change. Across eight scenarios, these behaviors worked reliably with minimal extra support and no pre-written fixes for specific situations.

Written by Kimi K2.5 using only the abstract above. Jev checked that every sentence is supported by the abstract sentences it cites, and removed any that were not. Point at a sentence to highlight its source.

Appeared: Thursday, October 22. bioRxiv. Preprint, not yet peer-reviewed.

DOI: 10.64898/2026.10.20.760214