La Agente Óptima: Towards Agentic Self-Driving Laboratories
Abstract
Self-driving laboratories (SDLs) combine automated experimentation with adaptive decision-making to accelerate scientific discovery. Their operation nevertheless often depends on human specialists who translate scientific objectives into executable closed-loop campaigns. Specialists adjust them as data and operating conditions change. Here, we present La Agente Óptima, an agentic framework that constructs and supervises Bayesian optimization campaigns across computational and experimental systems while maintaining a persistent optimization state. By separating large language model (LLM) reasoning from executed campaigns, Óptima runs repetitive optimization loops consistently, returns control to the agent only when progress requires interpretation or campaign revision, and keeps every decision auditable. We evaluate Óptima across ablation studies, five digital discovery tasks, and two physical platforms. Throughout, Óptima maintained executable campaigns as both the scientific problem and execution environment evolved. In a closed-loop contact angle optimization campaign, Óptima identified and corrected a mid-run measurement failure, bringing the contact angle from 71.4 to 67.8 degrees, just above the 64-66 degree range. From this result, Óptima correctly inferred that the target was likely unattainable with the available reagents and recommended changing the formulation. In a five-day multi-objective flow-chemistry campaign, Óptima increased the yield from 30% to 59% over 23 experiments. Despite substantial inference costs, it cost less and used substantially less starting material than a human-directed campaign, while selecting a more mass-efficient operating point. These results show that LLM-based agents can make rigorous, long-running optimization campaigns accessible to domain scientists without specialist setup, expanding the scope of SDLs.
Related
- What Can the Records of Published Self-Driving Laboratories Predict About a New Platform? A Step-Level Database and Explorer
- Lookahead Bayesian Optimisation for Molecular Growth in Self-Driving Laboratories
- Autonomous Robotic Workflow for Solution-Based Synthesis and High-Temperature Thermal Processing toward Self-Driving Laboratories
- Talking to instruments: device-integration standards for self-driving laboratories
- AEcroscopyWave: Towards Self-Driving Characterization Platforms for Agentic AI
BibTeX
@misc{muller2026agente,
title = {{La Agente Óptima: Towards Agentic Self-Driving Laboratories}},
author = {Marcel Müller and Jiaru Bai and Willi Gottstein and Abhijoy Mandal and Mohammad Nazeri and Elia Savino and Yanlin Fang and Sujoy Das and Sergio Pablo García Carrillo and Yeonghun Kang and Juan B. Pérez-Sánchez and Simone Pilon and Martin Fitzner and Timothy Noël and Frank Gu and Varinia Bernales and Alán Aspuru-Guzik},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.04564},
eprint = {2609.04564},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.04564},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - La Agente Óptima: Towards Agentic Self-Driving Laboratories
AU - Marcel Müller
AU - Jiaru Bai
AU - Willi Gottstein
AU - Abhijoy Mandal
AU - Mohammad Nazeri
AU - Elia Savino
AU - Yanlin Fang
AU - Sujoy Das
AU - Sergio Pablo García Carrillo
AU - Yeonghun Kang
AU - Juan B. Pérez-Sánchez
AU - Simone Pilon
AU - Martin Fitzner
AU - Timothy Noël
AU - Frank Gu
AU - Varinia Bernales
AU - Alán Aspuru-Guzik
PY - 2026
DA - 2026/09/03
PB - arXiv
DO - 10.48550/arxiv.2609.04564
UR - https://arxiv.org/abs/2609.04564
AB - Self-driving laboratories (SDLs) combine automated experimentation with adaptive decision-making to accelerate scientific discovery. Their operation nevertheless often depends on human specialists who translate scientific objectives into executable closed-loop campaigns. Specialists adjust them as data and operating conditions change. Here, we present La Agente Óptima, an agentic framework that constructs and supervises Bayesian optimization campaigns across computational and experimental systems while maintaining a persistent optimization state. By separating large language model (LLM) reasoning from executed campaigns, Óptima runs repetitive optimization loops consistently, returns control to the agent only when progress requires interpretation or campaign revision, and keeps every decision auditable. We evaluate Óptima across ablation studies, five digital discovery tasks, and two physical platforms. Throughout, Óptima maintained executable campaigns as both the scientific problem and execution environment evolved. In a closed-loop contact angle optimization campaign, Óptima identified and corrected a mid-run measurement failure, bringing the contact angle from 71.4 to 67.8 degrees, just above the 64-66 degree range. From this result, Óptima correctly inferred that the target was likely unattainable with the available reagents and recommended changing the formulation. In a five-day multi-objective flow-chemistry campaign, Óptima increased the yield from 30% to 59% over 23 experiments. Despite substantial inference costs, it cost less and used substantially less starting material than a human-directed campaign, while selecting a more mass-efficient operating point. These results show that LLM-based agents can make rigorous, long-running optimization campaigns accessible to domain scientists without specialist setup, expanding the scope of SDLs.
ER -