Automated Feedback Generation(AFG) has been proposed to give feedback automatically on assignments in a programming course. Existing AFG techniques generate feedback on incorrect submissions using reference programs such as the instructor’s answer program and correct submissions. In the case of the newly released assignments, however, taking advantage of the reference programs is quite challenging. To relieve the reference program dependency, existing AFG techniques focus on how to expand the reference programs or how to deal with the absence of reference programs for a specific assignment. However, even if the assignments are different, the ways to repair the novice student’s incorrect programs might be similar. In this paper, to verify whether our assumption is correct and to generate more repaired programs, we propose REFERENT, a transformer-based AFG technique using assignment information to secure the diversity and performance of feedback even when there is no available reference program. REFERENT performs transfer learning on the pre-trained model by securing data from the submission history of students. To generate assignment-related feedback, the goal is to generate guided feedback on the assignment by learning the title, tag, problem description, and test case as assignment information during training. We conducted a preliminary study to confirm the performance of REFERENT and the possibility of using assignment information. REFERENT was able to generate successful feedback for 32.7% of incorrect submissions without reference programs, and it was confirmed that the performance increased by up to 50.7% with reference programs. We also check the assignment information is useful for generating feedback