chunked/nltk_chunking/model_doc_rag/chunk_5.txt

However, their ability to access and precisely
manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind
task-specific architectures.