Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Commonsense QA

Benchmarks

Task NameDataset NameSOTA ResultTrend
Commonsense QACommonsense QA (ARC-E, ARC-C, HellaS, WinoG, BoolQ, OBQA, RTE, CoPa, Race) zero-shot
ARC-Easy Accuracy81.19
57
Commonsense Question AnsweringCommonsense QA
BoolQ Accuracy77.4
29
Commonsense Question AnsweringCommonsense QA
Reusability Score50.97
12
Question AnsweringCommonsense QA
PIQA80.85
12
Question AnsweringCommonsense QA
ARC-E Accuracy82.83
9
Question AnsweringCommonsense QA QA-avg
QA-avg Score61.67
4
Commonsense ReasoningCommonsense QA
Average Relative Improvement3.9
3
Commonsense Question AnsweringCommonsense QA 8 datasets
Average QA Score66.1
3
Showing 8 of 8 rows