Недавно вышло сразу три крутых датасета для бенчмаркинга Commonsense Reasoning
Event2Mind: Commonsense Inference on Events, Intents, and Reactions
https://homes.cs.washington.edu/~msap/pdfs/rashkin2018event2mind.pdf
COMET: Commonsense Transformersfor Automatic Knowledge Graph Construction
https://arxiv.org/pdf/1906.05317.pdf
Explain Yourself!Leveraging Language Models for Commonsense Reasoning
https://arxiv.org/pdf/1906.02361.pdf