Fetching the paper…

Do we really have to filter out random noise in pre-training data for language models? · Around