Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Healthcare through Large Language Models: A Study on Medical Question Answering

Haoran Yu, Chang Yu, Zihan Wang, Dongxian Zou, Hao Qin

TL;DR

This work investigates the deployment of large language models for medical question answering using the MedQuAD dataset. It compares multiple configurations and finds that the Sentence-t5 + Mistral 7B + Pretrain setup achieves the highest precision (0.762), driven by targeted pretraining, prompt construction, and domain-specific fine-tuning. The paper details a rigorous preprocessing pipeline, a hybrid architecture, and a sequence of training steps that leverage MLM and next-word objectives, along with techniques like learning rate scheduling and gradient clipping. The findings suggest that carefully engineered LLMs can deliver accurate medical information, potentially easing patient education and reducing clinician workload while highlighting areas for safety-focused deployment in real-world healthcare settings.

Abstract

In recent years, the application of Large Language Models (LLMs) in healthcare has shown significant promise in improving the accessibility and dissemination of medical knowledge. This paper presents a detailed study of various LLMs trained on the MedQuAD medical question-answering dataset, with a focus on identifying the most effective model for providing accurate medical information. Among the models tested, the Sentence-t5 combined with Mistral 7B demonstrated superior performance, achieving a precision score of 0.762. This model's enhanced capabilities are attributed to its advanced pretraining techniques, robust architecture, and effective prompt construction methodologies. By leveraging these strengths, the Sentence-t5 + Mistral 7B model excels in understanding and generating precise medical answers. Our findings highlight the potential of integrating sophisticated LLMs in medical contexts to facilitate efficient and accurate medical knowledge retrieval, thus significantly enhancing patient education and support.

Enhancing Healthcare through Large Language Models: A Study on Medical Question Answering

TL;DR

This work investigates the deployment of large language models for medical question answering using the MedQuAD dataset. It compares multiple configurations and finds that the Sentence-t5 + Mistral 7B + Pretrain setup achieves the highest precision (0.762), driven by targeted pretraining, prompt construction, and domain-specific fine-tuning. The paper details a rigorous preprocessing pipeline, a hybrid architecture, and a sequence of training steps that leverage MLM and next-word objectives, along with techniques like learning rate scheduling and gradient clipping. The findings suggest that carefully engineered LLMs can deliver accurate medical information, potentially easing patient education and reducing clinician workload while highlighting areas for safety-focused deployment in real-world healthcare settings.

Abstract

In recent years, the application of Large Language Models (LLMs) in healthcare has shown significant promise in improving the accessibility and dissemination of medical knowledge. This paper presents a detailed study of various LLMs trained on the MedQuAD medical question-answering dataset, with a focus on identifying the most effective model for providing accurate medical information. Among the models tested, the Sentence-t5 combined with Mistral 7B demonstrated superior performance, achieving a precision score of 0.762. This model's enhanced capabilities are attributed to its advanced pretraining techniques, robust architecture, and effective prompt construction methodologies. By leveraging these strengths, the Sentence-t5 + Mistral 7B model excels in understanding and generating precise medical answers. Our findings highlight the potential of integrating sophisticated LLMs in medical contexts to facilitate efficient and accurate medical knowledge retrieval, thus significantly enhancing patient education and support.
Paper Structure (32 sections, 11 equations, 1 figure, 1 table)

This paper contains 32 sections, 11 equations, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Dataset and Preprocessing
  4. Data Characteristics
  5. Data Preprocessing
  6. Data Cleaning
  7. Data Parsing
  8. Template Formatting
  9. Tokenization
  10. Data Augmentation
  11. Handling Class Imbalance
  12. Methodology
  13. Model Architectures
  14. Sentence-t5 Model
  15. Masked Language Model Objective
  16. ...and 17 more sections

Figures (1)

  • Figure 1: Overall architecture of the Sentence-t5 combined with Mistral 7B model.