Downstream Reasoning Tasks
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Downstream Reasoning Tasks ARC-c, ARC-e, BoolQ, HellaSwag, MMLU, OpenBookQA, PIQA, Winogrande
58.4ARC-c Accuracy (Zero-shot)
15
Downstream Reasoning Tasks (WikiText-2, ARC-e, ARC-c, BoolQ, PIQA, SIQA, HellaS., OBQA, Wino.)
11.78WikiText-2 Acc
6