This paper presents the development and evaluation of a Vietnamese Question Answer Generation (QAG) model specifically designed to generate review questions and exam questions in the Vietnamese subject Marxist-Lenininist Philosophy, helping learners to better manage their knowledge. Through the creation of a new QA dataset dedicated entirely to the Vietnamese subject Marxist-Lenininist Philosophy and the training and fine tuning processes, we have use a small-scale pre-trained language model (PLM) but the performance of generating Vietnamese questions is equivalent to that of LLM models. Our findings show that smaller specialized models can indeed match the capabilities of larger models, providing a more resource-efficient alternative without compromising on quality or efficiency. This study contributes to the broader discourse on the extensibility and adaptability of PLM, providing valuable insights into how to develop specialized models for specific educational purposes.