ResearchBenchmarksCommonsense Reasoning on OBQA (dev)Follow66.7AccuracyCKT61.18862.61964.0565.481Jun 4, 2023Evaluation ResultsMethodMethodLinksAccuracyCKTModel Scale=largeModel Scale=large2023.0666.7CALMModel Scale=largeModel Scale=large2023.0666T5Model Scale=largeModel Scale=large2023.0661.4