Visual Commonsense Reasoning On Vcr Q Ar Dev

Accuracy

Results

Performance results of various models on this benchmark

Comparison Table

Model Name	Accuracy
pevl-position-enhanced-pre-training-and	57.8