One of the Approaches to Natural Language Prompting for the Large Language Model Using the University of Tyumen Document Storage

The following paper explores the application of Retrieval augmented generation (RAG) for answering questions posed by students at the University of Tyumen using the large language model (LLM) and the corporate wiki system. To enhance the precision of document retrieval for question-answering purposes, the sentence embedding model is fine-tuned on the set of questions that has been generated by the LLM to corporate university documents. The results demonstrate improved performance compared to using embeddings from a third party LLM service. As a result the microservice for answering students’ questions has been developed and tested for using it throughout the UTMN Virtual Assistant.