<p>This study explores the dynamic evolution of the Qing dynasty’s imperial examination system, focusing on emperors’ decision-making attitudes and examination-related events. Using an LLM (ChatGLM-4) to extract information from historical documents and a pre-trained embedding model (Embedding-2) to quantify text, combined with classification machine learning to predict event nature and decision-making attitudes, this study provides insights into the decision dynamics of the examination system. ChatGLM-4 annotated sample labels and feature importance analysis identified critical factors affecting decision-making attitudes and event nature. Key findings include: (a) temporal variations in event distribution; (b) historical context and event elements (theme, cause, suggestion) primarily influence decision-making attitudes, while decision-making attitudes and elements (theme, decision, suggestion) primarily affect event nature; (c) the emperor’s age shows a neutral-negative-positive influence on decision-making attitudes with increasing age, though individual emperors had a relatively minor impact overall. The conclusions highlight the imperial examination system’s stability, fairness, and adaptability, though its inherent defects limited its fairness and adaptability, leading to its decline.</p>