Abstract: Based on a domestic mainstream large language model, we designed a question answering system for the Linux course. This system, combined with  retrieval enhancement technology, could continuously learn from human feedback, which helped to solve the problem of how to more effectively assist students’ learning in the Linux course teaching. Experimental results show that the system improves the factuality of answers provided by the large language model and can effectively answer  students’ questions. In addition, the system accumulates a professional domain knowledge base presented in the form of natural language at a  lower cost, reducing the workload of teachers in collecting and organizing teaching materials.

Key words: Linux course, large language model, continuous learning, question answering system