Do different prompting methods yield a common task representation in language models?

Re-assign community