Broaden the Vision: Geo-Diverse Visual Commonsense Reasoning

Benchmark Model Rank Results
visual-commonsense-reasoning-on-gd-vcrHuman#1Accuracy: 88.84
visual-commonsense-reasoning-on-gd-vcrViLBERT#2Accuracy: 59.99Gap (West): -7.28
visual-commonsense-reasoning-on-gd-vcrVisualBERT#3Accuracy: 53.95Gap (West): -10.42
visual-commonsense-reasoning-on-gd-vcrText-only BERT#4Accuracy: 35.33