This dataset contains chunked extracts (of ~300 tokens) from papers related to (and including) the Llama 2 research paper. Related papers were identified by following a trail of references, extracting those papers with the arxiv-bot package, and repeating.
