The large language models established based on open AI have shown great potential in improving the efficiency of knowledge acquisition,which also brings challenges to traditional teaching,learning and assessment methods.Compared with undergraduates,graduate students need to face the frontier of engineering and aim to the academic frontier.Consequently,their training goals are more focused on the comprehensive application of knowledge,and the training method is relatively moreopen.Based on the teaching practice of"Soil Mechanics"course for graduate students in the major of traffic engineering,this paper discusses the featuredcontents,teaching methods,teacher-student interaction and measures around the fundamental task of foster virtue.The discussions included in this paper consider the needs for the development situations of AI and graduate student education,and meanwhile focusing on the cultivation of people and education development goals in the new era.