Commonsense QA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Commonsense QA (ARC-E, ARC-C, HellaS, WinoG, BoolQ, OBQA, RTE, CoPa, Race) zero-shot
81.19ARC-Easy Accuracy
57
Commonsense QA
77.4BoolQ Accuracy
29
Commonsense QA
50.97Reusability Score
12
Commonsense QA
80.85PIQA
12
Commonsense QA
82.83ARC-E Accuracy
9
Commonsense QA QA-avg
61.67QA-avg Score
4
Commonsense QA
3.9Average Relative Improvement
3
Commonsense QA 8 datasets
66.1Average QA Score
3