Commonsense Question Answering on CosQA (test)
82AccuracyFew-shot Accuracy
Evaluation Results
| Method | Links | |
|---|---|---|
| Few-shot AccuracyMode=Few-shot2026.03 | 82 | |
| Zero-shot AccuracyMode=Zero-shot2026.03 | 81.7 | |
| Linear ProbeTrain Dataset=CosQA2026.03 | 81.64 | |
| TaTTrain Dataset=CosQA2026.03 | 80.94 | |
| TaTTrain Dataset=OpenQA2026.03 | 76.42 | |
| TaTTrain Dataset=ARC-E2026.03 | 75.08 | |
| TaTTrain Dataset=ARC-C2026.03 | 74.24 | |
| Linear ProbeTrain Dataset=ComQA2026.03 | 74.18 | |
| Linear ProbeTrain Dataset=ARC-C2026.03 | 73.34 | |
| Linear ProbeTrain Dataset=Hellaswag2026.03 | 71.17 | |
| Linear ProbeTrain Dataset=ARC-E2026.03 | 66.83 | |
| TaTTrain Dataset=Hellaswag2026.03 | 66.4 | |
| TaTTrain Dataset=ComQA2026.03 | 64.89 | |
| Linear ProbeTrain Dataset=SiQA2026.03 | 61.54 | |
| Linear ProbeTrain Dataset=OpenQA2026.03 | 61.53 | |
| Linear ProbeTrain Dataset=BoolQ2026.03 | 61.51 | |
| TaTTrain Dataset=SiQA2026.03 | 59.73 | |
| TaTTrain Dataset=BoolQ2026.03 | 51.56 |