Fine-Tuning Large Language Model (LLM) to Answer Basic Questions for Prospective New Students at Syiah Kuala University Using the Retrieval-Augmented Generation (RAG) Method

USK Mistral 7B is a large language model designed to answer basic admission questions at Universitas Syiah Kuala (USK). The model was fine-tuned using the open-source model of Mistral 7B using collected data from admissions and lectures at the university. The QLoRA and RAG techniques were used to train the model and retrieve relevant information from external data sources. The results were evaluated using the ROUGE score. Responses were generated with a score of >0.5 on ten out of 46 questions with the RAG method, and testing with the fine-tuning method was carried out on 20 questions and resulted in responses with a score of 1.0 from all questions asked. The performance of USK Mistral 7B shows its potential as an effective tool in helping students querying information about admission and lectures at USK.