Can large language models generate salient negative statements?

Re-assign community