Leveraging Retrieval-Augmented Generation In Large Language Models for Effective Learning: A Data Structures & Algorithms Learning Assistant

Abstract

In the rapidly evolving field of artificial intelligence, integrating Retrieval-Augmented Generation (RAG) in Large Language Models (LLMs) provides an innovative approach to reshaping education. This paper explores the design and development of a learning assistant based on an LLM, specifically for Data Structures and Algorithms (DSA), using the RAG technique. According to [1], DSA is a core course that students need help grasping due to the abstractness of its concepts. With the advent of LLMs, students are now adopting these tools to support their learning [2]. However, LLMs have been shown to exhibit some issues of bias, hallucinations and generating irrelevant output, which may not be helpful to the students [3]. Therefore, this research seeks to address these problems using RAG to tailor the LLM's responses and better support students in learning data structures and algorithms. It seeks to answer the question: "How can RAG be leveraged to develop an effective DSA learning assistant?" The technology acceptance model and task technology fit provide the theoretical frameworks for evaluating the effectiveness and usefulness of the learning assistant. Previous research has demonstrated the value of RAG in customising LLMs for different domains; however, research has yet to be conducted on its application to data structures and algorithms. Therefore, further research is needed to bridge this gap and contribute to the ongoing revolution of AI in education. Results show that the RAG technique is effective in performing the task of generating accurate and relevant content.