<p>The article examines the integration of retrieval-augmented generation (RAG) technology into formative assessment in computer science teaching. Traditional automatic short answer grading (ASAG) methods based only on text similarity do not allow for a deep analysis of student responses in terms of content and context. To solve this problem, a hybrid model combining RAG and ASAG technologies is proposed. As a result of the proposed assessment criteria, the model evaluates student responses not only in terms of language, but also on the basis of computer science knowledge, teaching materials, and expert examples. This approach allows for a more objective and transparent assessment of students’ knowledge application, critical thinking, and creativity skills. The research results show that the RAG-based hybrid model provides personalized feedback, facilitates monitoring of the learning process at both the individual and group levels, and develops students’ self-directed learning skills.</p>