List your company

ResearchFormerly WizWand

BigBench-Hard

Benchmarks

TaskDatasetSOTA resultTrendResults

Logical Reasoning

BigBench Hard Boolean Expressions

76.8Accuracy

17

BigBench Hard Penguins

44.1Accuracy

5

Linguistic Reasoning

BigBench Hard Disambiguation QA

55.1Accuracy

5

BigBench-Hard collection averaged

45.41Ours Accuracy

4