Pretrained models are powerful tools for natural language processing, but they often lack the ability to answer basic questions about themselves.