ResearchBenchmarksCommonsense Reasoning on Winogrande v1.1 (test)Follow91.28AccuracyUNICORN69.356875.048480.7486.4316Jan 28, 2022Evaluation ResultsMethodMethodLinksAccuracyUNICORNsetting=Supervisedsetting=Supervised2022.0191.28MT-NLGsetting=Few-shotsetting=Few-shot2022.0178.85GPT-3setting=Few-shotsetting=Few-shot2022.0177.7MT-NLGsetting=One-shotsetting=One-shot2022.0173.72GPT-3setting=One-shotsetting=One-shot2022.0173.2MT-NLGsetting=Zero-shotsetting=Zero-shot2022.0173.01GPT-3setting=Zero-shotsetting=Zero-shot2022.0170.2Gophersetting=Zero-shotsetting=Zero-shot2022.0170.2