Authors

Yuanzhe Jin, Jiali Yu

Abstract

As more and more countries pay attention to the quality of university education in recent years, the number of students and mentors in schools has grown rapidly. In the higher education scenario, the communication between mentors and students has long been a focal point of research. Within the communication between students and mentors, understanding information in non-face-to-face interactions often poses a challenge to communication efficiency. When using instant messaging tools for text exchanges, recipients often find themselves needing to infer the sender’s intentions. With the recent rapid development of large language models (LLMs) in text comprehension capabilities, we propose to use LLMs to encode information states of the sending message, aiming to achieve a more accurate understanding of the states between mentors and students. Through experiments, we have found that using LLMs for encoding information states can assist mentors and students in better understanding responses to each other and avoid misunderstandings.

Download paper here