Luo Feifei, Zhang Jinglang, Wang Qilong, Yang Chunpeng
Tianjin Key Laboratory of Advanced Carbon and Electrochemical Energy Storage, School of Chemical Engineering and Technology, Tianjin University, Tianjin 300350, China.
Tianjin Key Lab of Machine Learning, College of Intelligence and Computing, Tianjin University, Tianjin 300350, China.
ACS Cent Sci. 2025 Apr 2;11(4):511-519. doi: 10.1021/acscentsci.4c01935. eCollection 2025 Apr 23.
Artificial intelligence (AI) using large language models (LLMs) such as GPTs has revolutionized various fields. Recently, LLMs have also made inroads in chemical research even for users without expertise in coding. However, applying LLMs directly may lead to "hallucinations", where the model generates unreliable or inaccurate information and is further exacerbated by limited data set and inherent complexity of chemical reports. To counteract this, researchers have suggested prompt engineering, which can convey human ideas formatively and unambiguously to LLMs and simultaneously improve LLMs' reasoning capability. So far, prompt engineering remains underutilized in chemistry, with many chemists barely acquainted with its principle and techniques. In this Outlook, we delve into various prompt engineering techniques and illustrate relevant examples for extensive research from metal-organic frameworks and fast-charging batteries to autonomous experiments. We also elucidate the current limitations of prompt engineering with LLMs such as incomplete or biased outcomes and constraints imposed by closed-source limitations. Although LLM-assisted chemical research is still in its early stages, the application of prompt engineering will significantly enhance accuracy and reliability, thereby accelerating chemical research.
使用诸如GPT等大语言模型（LLM）的人工智能（AI）已经彻底改变了各个领域。最近，即使对于没有编码专业知识的用户，LLM也已进入化学研究领域。然而，直接应用LLM可能会导致“幻觉”，即模型生成不可靠或不准确的信息，而且有限的数据集和化学报告固有的复杂性会进一步加剧这种情况。为了应对这一问题，研究人员提出了提示工程，它可以将人类的想法以一种有建设性且明确的方式传达给LLM，同时提高LLM的推理能力。到目前为止，提示工程在化学领域的应用仍然不足，许多化学家对其原理和技术几乎一无所知。在这篇展望文章中，我们深入探讨了各种提示工程技术，并举例说明了从金属有机框架、快速充电电池到自主实验等广泛研究中的相关示例。我们还阐明了当前使用LLM进行提示工程的局限性，例如结果不完整或有偏差以及闭源限制所带来的约束。尽管LLM辅助的化学研究仍处于早期阶段，但提示工程的应用将显著提高准确性和可靠性，从而加速化学研究。
