Do Large Language Models Know about Facts?