Enhancing elusive clues in knowledge learning by contrasting attention of language models

Open in new window