chunked/nltk_chunking/model_doc_rembert/chunk_1.txt

The abstract from the paper is the following:
We re-evaluate the standard practice of sharing weights between input and output embeddings in state-of-the-art
pre-trained language models.