NLU Suite
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
NLU Suite (MMLU, SST-2, AGNews, 20News, MNLI, SNLI)
89.9Average Accuracy
31
NLU suite Zero-Shot (CSQA, SIQA, HS, WG, PIQA, OBQA, ARC:E, ARC:C)
49.8CSQA Accuracy
8
NLU Suite NLI
0.754AUC
4