Commonsense
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Commonsense 8 Sub-Tasks
88.22Accuracy (8 Sub-Tasks)
80
Commonsense170k (test)
76.4BoolQ Accuracy
40
Commonsense
74.6RCC
29
Commonsense
10Mean Improvement
9
Commonsense-15K (test)
36.11ARC-Challenge Accuracy
7
Commonsense-15K
53.33ARC-Challenge Accuracy
5
Commonsense Gender (test)
20Accuracy
5
Commonsense Race (test)
40.4Correctness Rate
5
CommonSense
25.969Perplexity
3