Commonsense Reasoning on CommonsenseQA 3k (dev)
71.52Accuracy (5%)T5-3B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| T5-3BAugmentation=VAWI-SBS+PET2022.12 | 71.52 | 74.19 | 76.49 | 83.61 | |
| T5-3BAugmentation=+Images2022.12 | 70.96 | 73.6 | 75.91 | 82.4 | |
| T5-3BAugmentation=+None2022.12 | 70.16 | 73.02 | 75.04 | 81.81 | |
| RoBERTa-largeAugmentation=+VAWI-SBS2022.12 | 50.27 | 58.17 | 62.22 | 76.54 | |
| RoBERTa-largeAugmentation=+Images2022.12 | 49.55 | 57.78 | 61.29 | 75.61 | |
| RoBERTa-largeAugmentation=+None2022.12 | 48.39 | 56.3 | 59.06 | 74.19 | |
| RoBERTa-baseAugmentation=+VAWI-SBS2022.12 | 42.94 | 49.27 | 53.97 | 65.1 | |
| RoBERTa-baseAugmentation=+Images2022.12 | 42.37 | 48.09 | 52.81 | 64.22 | |
| RoBERTa-baseAugmentation=+None2022.12 | 41.88 | 46.04 | 50.58 | 61.88 |