Sovereign 8s.is mirror of standardgalactic/llms-in-clinical-medicine-systematic-review
  • Jupyter Notebook 84.7%
  • JavaScript 12%
  • CSS 3.3%
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2025-11-28 12:38:21 -08:00
Batch Responses Initial commit 2025-10-19 18:26:31 -04:00
Data Extraction Helpers Initial commit 2025-10-19 18:26:31 -04:00
Human Screening and Tiering Data Initial commit 2025-10-19 18:26:31 -04:00
Prompts Initial commit 2025-10-19 18:26:31 -04:00
standalone-web Initial commit 2025-10-19 18:26:31 -04:00
analysis-and-figures.ipynb Changed stat-test in notebook to match paper 2025-11-28 12:38:21 -08:00
compiled_deduped_with_abstracts.csv Initial commit 2025-10-19 18:26:31 -04:00
deduped_and_processed_studies-GPT-5r-high.csv Initial commit 2025-10-19 18:26:31 -04:00
deduped_and_processed_studies.jsonl Initial commit 2025-10-19 18:26:31 -04:00
embase-export-9-6-25.csv Initial commit 2025-10-19 18:26:31 -04:00
extracted_data_parsing_dataset.jsonl Initial commit 2025-10-19 18:26:31 -04:00
final_processed_studies.jsonl Initial commit 2025-10-19 18:26:31 -04:00
LICENSE.md Initial commit 2025-10-19 18:26:31 -04:00
pubmed-export-9-6-25.csv Initial commit 2025-10-19 18:26:31 -04:00
README.md Initial commit 2025-10-19 18:26:31 -04:00
scopus-export-9-6-25.csv Initial commit 2025-10-19 18:26:31 -04:00
screening.ipynb Initial commit 2025-10-19 18:26:31 -04:00

About

This repository contains the associated code and data necessary to reproduce the results of our systematic review of LLMs in clinical medicine. Additionally, this notebook can be modified and built upon to conduct your own systematic review as well!

screening.ipynb contains the code that both runs the LLM screening process and analyzes the results against human ground truth data.

analysis-and-figures.ipynb contains the code that produces all of the analysis and figures after the screening has been conducted.

standalone-web/ contains the HTML/JS/CSS code that allows you to directly browse the generated data. No extra installation or libraries are needed here; just double-click click_me.html to launch!

Please reach out if you have any questions or comments!