Most of them are experimental studies. So it would be text extraction of something like title, authors, species of the study, sample size etc. And classify based on the content of the pdfs.

I tried the GPT-4o, it's good but it'll cost a lot if I want to process all the documents.