Visual Commonsense Reasoning on VCR
79.8Q->A AccuracyUNITER-large
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| UNITER-largeensemble_size=102019.09 | 79.8 | 83.4 | 66.8 | |
| UNITER-largeensemble_size=single2019.09 | 77.3 | 80.8 | 62.8 | |
| ViLBERTensemble_size=102019.09 | 76.4 | 78 | 59.8 | |
| VLBERT-largeensemble_size=single2019.09 | 75.8 | 78.4 | 59.7 |