A factuality benchmark called SimpleQA that measures the ability for language models to answer short, fact-seeking questions.
