share benchmark results in the docs with the dataset viewer