ResearchBenchmarksCommonsense Reasoning on CoS-E v1.0Follow0.695Score (Finetune+Baseline vs Predict+Baseline)T5-base0.504680.554090.60350.65291May 4, 2023Evaluation ResultsMethodMethodLinksScore (Finetune+Baseline vs Predict+Baseline)Score (Finetune+Baseline vs Predict+Infusion)Simulatability ScoreScore (Finetune+Infusion vs Predict+Infusion)TREU ScoreT5-baseBackbone=T5-baseBackbone=T5-base2023.050.6950.645-0.050.8780.133BART-baseBackbone=BART-baseBackbone=BART-base2023.050.5120.486-0.0260.790.252