Advancing Question-Answering in Ophthalmology With Retrieval-Augmented Generation: Benchmarking Open-Source and Proprietary Large Language Models
Translational Vision Science & Technology · 2025
DOI: 10.1167/tvst.14.9.18 ↗Kurz gefasst
Zurückgezogene Publikation in der Zeitschrift Translational Vision Science & Technology, 2025. Verfasst von Quang Nguyen und 11 weiteren Autor:innen. Beteiligte Einrichtung: UCL Institute of Ophthalmology, London, UK. Hauptthemen laut MEDLINE: Benchmarking, Ophthalmology, Information Storage and Retrieval. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Purpose
The purpose of this study was to evaluate the application of combining information retrieval with text generation using Retrieval-Augmented Generation (RAG) to benchmark the performance of open-source and proprietary generative large language models (LLMs) in question-answering in ophthalmology.
Methods
Our dataset comprised 260 multiple-choice questions sourced from two question-answer banks designed to assess ophthalmic knowledge: the American Academy of Ophthalmology's (AAO) Basic and Clinical Science Course (BCSC) Self-Assessment program and OphthoQuestions. Our RAG pipeline retrieves documents in the BCSC companion textbook using ChromaDB, followed by reranking with Cohere to refine the context provided to the LLMs. Generative Pretrained Transformer (GPT)-4-turbo and 3 open-source models (Llama-3-70B, Gemma-2-27B, and Mixtral-8 × 7B) are benchmarked using zero-shot, zero-shot with Chain-of-Thought (zero-shot-CoT), and RAG. Model performance is evaluated using accuracy on the two datasets. Quantization is applied to improve the efficiency of the open-source models. Effects of quantization level are also measured.
Results
Using RAG, GPT-4-turbo's accuracy increased by 11.54% on BCSC and by 10.96% on OphthoQuestions. Importantly, the RAG pipeline greatly enhances overall performance of Llama-3 by 23.85%, Gemma-2 by 17.11%, and Mixtral-8 × 7B by 22.11%. Zero-shot-CoT had overall no significant improvement on the models' performance. Quantization using 4 bit was shown to be as effective as using 8 bits while requiring half the resources.
Conclusions
Our work demonstrates that integrating RAG significantly enhances LLM accuracy especially for smaller LLMs. Translation Relevance: Using our RAG, smaller privacy-preserving open-source LLMs can be run in sensitive and resource-constrained environments, such as within hospitals, offering a viable alternative to cloud-based LLMs like GPT-4-turbo.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- UCL Institute of Ophthalmology, London, UK.
- UCL Institute of Health Informatics, London, UK.
- Moorfields Eye Hospital NHS Foundation Trust, London, UK.
- CT MedTech Ltd, Ho Chi Minh City, Vietnam.
- Department of Informatics, New Jersey Institute of Technology, Newark, NJ, USA.
- Stanford Byers Eye Institute, Palo Alto, CA, USA.
Zitieren
APA
Nguyen, Q., Nguyen, D. A., Dang, K., Liu, S., Wang, S. Y., Woof, W. A., Thomas, P. B. M., Patel, P. J., Balaskas, K., Thygesen, J. H., Wu, H., & Pontikos, N. (2025). Advancing Question-Answering in Ophthalmology With Retrieval-Augmented Generation: Benchmarking Open-Source and Proprietary Large Language Models. Translational Vision Science & Technology. https://doi.org/10.1167/tvst.14.9.18
Vancouver
Nguyen Q, Nguyen DA, Dang K, Liu S, Wang SY, Woof WA, et al. Advancing Question-Answering in Ophthalmology With Retrieval-Augmented Generation: Benchmarking Open-Source and Proprietary Large Language Models. Translational Vision Science & Technology. 2025. doi:10.1167/tvst.14.9.18
BibTeX
@article{Nguyen2025,
author = {Nguyen, Quang and Nguyen, Duy-Anh and Dang, Khang and Liu, Siyin and Wang, Sophia Y and Woof, William A and Thomas, Peter B M and Patel, Praveen J and Balaskas, Konstantinos and Thygesen, Johan H and Wu, Honghan and Pontikos, Nikolas},
title = {Advancing Question-Answering in Ophthalmology With Retrieval-Augmented Generation: Benchmarking Open-Source and Proprietary Large Language Models},
journal = {Translational Vision Science & Technology},
year = {2025},
doi = {10.1167/tvst.14.9.18}
}
EUVIMED