Using RAG (Retrieval Augmented Generation) and “ground truth” knowledge bases (in this case, Pubmed papers and curated articles) to find cures for rare diseases.

CASE STUDY: How AI Can Help Cure Rare Diseases

3 Likes

Are you keeping track of which origin of the data is contributing most to the valid results?

I am curious if the PubMed PDFs contribute most of the valid results.

For each response, the origin of the chunks is used to calculate the citations (in the citation calculation algorithm – based on the passed in chunks and the response from GPT-4)

That part I am not sure about – because it will depend on whatever queries users are doing (that’s more of an analytics thing I guess)

1 Like