Fig. 6: LLM-based agents facilitated self-driven reaction condition optimization.
From: An automatic end-to-end chemical synthesis development platform powered by large language models
a The large language model (LLM)-based agents copiloted self-driven reaction optimization system. Users interface with the hardware system via natural language through the web application with Experiment Designer and Hardware Executor as the backend. The exact transcript of the natural language description of the task is provided in Supplementary Table 37-38. The automated reaction optimization platform, driven by Bayesian optimization (BO) algorithm, performed closed-loop reaction and analysis using automated Unchained synthesis platform and high-performance liquid chromatography (HPLC), respectively. The image of Unchained Labs Big Kahuna synthesis platform was obtained from the Unchained Labs website (https://unchained-labs.cn). The evolution profile of (b) yield and (c) probability of improvement (PI) value during the closed-loop reaction optimization process. d Result Interpreter’s recommendations on whether reaction optimization should be terminated at 6th, 12th, 22nd, and 26th experiment (see detailed interaction dialogs in Supplementary Table 40).
