Publications by authors named "Sandeep Vemulapalli"
Comparison of Prompt Engineering and Fine-Tuning Strategies in Large Language Models in the Classification of Clinical Notes.
AMIA Jt Summits Transl Sci Proc
May 2024
Article Synopsis
- - The study evaluates the effectiveness of large language models (LLMs), particularly GPT-3.5 Turbo, GPT-4, and Llama-7B, in identifying metastatic cancer from medical discharge summaries, compared to BERT models and medical expert annotations.
- - Results showed that well-structured prompts with reasoning steps significantly improved the models' performance, with GPT-4 outperforming all other models in the study.
- - The research indicates that GPT-4 could replace specialized models like PubMedBERT in clinical settings due to its strong performance, even without reliance on specific keywords or fine-tuning enhancements.
Article Synopsis
- - The study evaluates large language models (LLMs), including GPT-3.5 Turbo and GPT-4, in the context of healthcare, specifically for identifying metastatic cancer from discharge summaries.
- - Results show that effective prompt engineering significantly improves model performance, with GPT-4 outperforming other models, while methods like one-shot learning and fine-tuning did not provide additional benefits.
- - The findings indicate that GPT-4 could potentially replace specialized models like PubMedBERT using strategic prompts, and highlight the need for enhancements in open-source models to better fit clinical applications.