chunked/nltk_chunking/model_doc_byt5/chunk_4.txt

Since byte or character sequences are longer than token
sequences, past work on token-free models has often introduced new model architectures designed to amortize the cost of
operating directly on raw text.