chunked/nltk_chunking/model_doc_blenderbot/chunk_2.txt

While prior work has shown that
scaling neural models in the number of parameters and the size of the data they are trained on gives improved results,
we show that other ingredients are important for a high-performing chatbot.