Incorporating Large Language Models (LLMs) into Modern Neuroscience

---

Background

Large Language Models (LLMs) are generative AI models that are able to produce natural language text interactively with users. By interacting with LLMs in a programmatic way, there are a variety of data summarization and extraction processes that can be reliably automated. We are exploring possibilities of extracting study specific data for conducting meta-analyses.

In general, we are looking to establish expertise working with this increasingly popular and transformative technology, so we can understand how to safely and ethically incorporate it into our work.

Projects

• reviewer2go: a series of prompts and queries for producing a simple report of the COBIDAS criteria from a published paper.
• litmining ecosystem: A set of accessible, scalable, and reliable tools for mining the biomedical literature.
• biomed-llm-retrieval: This repo is built by our collaborator Alejandro de la Vega (U. Texis at Austin); it uses LLMs to extract annotations from the abstract/texts of biomedical papers, with a focus on fMRI.

People

Brent McPherson

Postdoctoral researcher

Kendra Oudyk

Postdoctoral researcher

Alice Chen

PhD rotation student