Evaluation script for the GQA dataset. 
eval.py computes metrics such as accuracy, consistency, plausibility, grounding scores etc.
choices.json files are supporting files for the evaluation.

Pleae visit gqadataset.org for all information about the dataset, including examples, visualizations, paper and slides.
