Article 2 Prompt Phrasing Shifts Model Performance More Than Tier β and Which LLMs Will Get Your Jokes