Can Prompt Probe Pretrained Language Models? Understanding the Invisible Risks from a Causal View

Boxi Cao, Hongyu Lin, Xianpei Han, Fangchao Liu, Le Sun