Identifying plausible explanations on WinoWhy
87.55AccuracyRoBERTa-Grande
Evaluation Results
| Method | Links | |
|---|---|---|
| RoBERTa-Grandeknowledge-augmented=true, plausibility annotations=with2023.05 | 87.55 | |
| RoBERTaplausibility annotations=with2023.05 | 75.04 | |
| QNLI-ATOMIC (N)knowledge-augmented=true, plausibility annotations=without2023.05 | 73.47 | |
| LiPoRplausibility annotations=without2023.05 | 69.88 | |
| ZS GPT-NEOsetting=zero-shot, plausibility annotations=without2023.05 | 59.13 | |
| Previous Bestplausibility annotations=without2023.05 | 56.37 | |
| Tuned BARTsetting=tuned, plausibility annotations=without2023.05 | 55.32 | |
| ZS GPT3setting=zero-shot, plausibility annotations=without2023.05 | 50.99 | |
| ZS BARTsetting=zero-shot, plausibility annotations=without2023.05 | 45.69 | |
| KDDC-CSKG (N)knowledge-augmented=true, plausibility annotations=without2023.05 | 44.05 | |
| KDDC-ATOMIC (N)knowledge-augmented=true, plausibility annotations=without2023.05 | 42.44 | |
| KDDC-CWWV (N)knowledge-augmented=true, plausibility annotations=without2023.05 | 40.45 |