MedQAS: A Medical Question Answering System Based on Finetuning Large Language Models

Large language models (LLMs) are transforming the field of natural language processing, offering remarkable capabilities in text generation, language understanding, and conversational Artificial Intelligence. However, their size and complexity often hinder their deployment in resource-constrained environments. To address these concerns, we leverage large-scale instruction-following datasets from open-source projects. We then apply parameter-efficient tuning using Low-Rank Adaptation (LoRA) on two open Large Language Models (LLMs): Bloomz (Multilingual) and LLaMA. Finally, we evaluate the effectiveness of our approach on a per-sample basis, considering the helpfulness, relevance, accuracy, and level of detail in their responses. This evaluation process involves using GPT-4 as an automated scoring mechanism. Despite utilizing a low-cost setup, our proposed Medical Question Answering System (MedQAS) shows improvement over the original models in our evaluation tasks.