大语言模型时代背景下数据库课程实验考核改革研究

Research on the Reform of Laboratory Assessment in Database Courses in the Era of Large Language Models

摘要: 针对大语言模型背景下数据库课程实验考核以结果正确为唯一依据、难以甄别学生真实编程能力的问题，文章剖析了传统实验考核模式失效的表现与成因，提出由过程证据留痕、现场代码答辩、抗模型化命题与达成度数据反哺构成的实验考核改革方案，阐述了如何将评价重心由结果验收转向过程可信与能力可证，并对改革后的考核环节及其权重构成进行系统设计，最后通过现场考察、课业成绩与问卷调查三个方面说明教学实践效果。

Abstract: To address the problem that laboratory assessment in database courses relies solely on the correctness of results and can hardly discriminate students’ authentic programming ability in the era of large language models, this paper analyzes the manifestations and causes of the failure of the traditional laboratory assessment mode, and proposes a reform scheme composed of the retention of process evidence, on-site oral defense of code, anti-model question design, and feedback from achievement data. It expounds how to shift the focus of assessment from result acceptance to credible process and demonstrable competence, systematically designs the reformed assessment components and their weightings, and finally illustrates the effect of the teaching practice through on-site examination, course performance, and questionnaire results.