Task Suite
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Zero-shot Task Suite (HellaSwag, MathQA, MMLU, OpenBookQA, WinoGrande, GSM8K, HumanEval)
82.72HellaSwag Accuracy
31
Task Suite Zero-shot (ARC-e, ARC-c, HellaSwag, OBQA, WinoGrande, MathQA, PIQA)
83.54ARC-e
17