ARC, HellaSwag, PIQA, WinoGrande
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ARC-Easy, ARC-Challenge, HellaSwag, PIQA, WinoGrande lm-evaluation-harness (test)
82.87ARC-e Accuracy
43
ARC-C, ARC-E, HellaSwag, PIQA, and WinoGrande
74.35Downstream Average
10