【摘要】 基于结构坐标的传统物质科学理论局限在对少量维度、静态结构、微观细节的研究中，难以处理高维度、动态多变的复杂化学体系。为了发展不依赖结构坐标的物质科学新理论，并基于该理论发展用于化学创制的新手段，国家自然科学基金委员会成功举办了第374期双清论坛“基于谱学大模型的物质科学认知”。本文旨在总结论坛达成的共识，分析了近年来谱学与人工智能结合领域在国际国内的研究现状和动态，重点讨论了当前该领域研究所面临的重大机遇和挑战，凝练了该领域未来5～10年的重大关键科学问题和核心技术，建议在原位谱学数据库与标准化建设、谱学数据解析与智能算法发展、谱学大模型构建与优化、跨学科的谱学大模型应用以及基于谱学大模型的高效材料研发平台等相关领域重点攻关，并就科学基金重点资助项目给出了建议。更多还原

【Abstract】 The traditional substance science theories based on structural coordinates are confined to studying limited dimensions,static structures,and microscopic details,making it difficult to handle high-dimensional,dynamically complex chemical systems. To develop new substance science theories that do not rely on structural coordinates and to establish novel approaches for chemical innovation,the National Natural Science Foundation of China successfully organized the374^thShuangqing Forum titled “Cognitive Material Science Based on Spectral Big Models”. This article aims to summarize the consensus reached at the forum and analyze the recent research progress and trend in the field of integrating spectroscopy and artificial intelligence domestically and abroad. It focuses on the significant opportunities and challenges currently faced in this field. The paper distills the key scientific frontier and core technologies for the future 5～10 years in this field,such as establishing standardized in-situ spectroscopic database,development of intelligent algorithm for spectroscopic data analysis,construction and application of spectroscopic large model,and development of efficient substance research platform driven by spectroscopic large models. In addition,the potential scientific research funding strategy is also suggested.更多还原