List your company

Commonsense Question Answering on CosQA (test)

82AccuracyFew-shot Accuracy
50.342458.561266.7874.9988Mar 1, 2026

Evaluation results

MethodLinks
Few-shot AccuracyMode=Few-shot2026.0382
Zero-shot AccuracyMode=Zero-shot2026.0381.7
Linear ProbeTrain Dataset=CosQA2026.0381.64
TaTTrain Dataset=CosQA2026.0380.94
TaTTrain Dataset=OpenQA2026.0376.42
TaTTrain Dataset=ARC-E2026.0375.08
TaTTrain Dataset=ARC-C2026.0374.24
Linear ProbeTrain Dataset=ComQA2026.0374.18
Linear ProbeTrain Dataset=ARC-C2026.0373.34
Linear ProbeTrain Dataset=Hellaswag2026.0371.17
Linear ProbeTrain Dataset=ARC-E2026.0366.83
TaTTrain Dataset=Hellaswag2026.0366.4
TaTTrain Dataset=ComQA2026.0364.89
Linear ProbeTrain Dataset=SiQA2026.0361.54
Linear ProbeTrain Dataset=OpenQA2026.0361.53
Linear ProbeTrain Dataset=BoolQ2026.0361.51
TaTTrain Dataset=SiQA2026.0359.73
TaTTrain Dataset=BoolQ2026.0351.56