Code for the benchmark containing dataset, models and metrics for productive concept learning -- a kind of compositional reasoning task that requires reasoning about uncertainty and learning compositionally rich and challenging concepts in a low-shot, meta-learning framework. - View it on GitHub