I found this project recently which is really interesting: https://alex.macrocosm.so/download

They've embedded all papers on Arxiv and made the results available for anyone to use.

I believe they used the Instructor XL model - I need to build an LLM plugin for that.