Just read twice: closing the recall gap for recurrent language models

Open in new window