These targets allow one experiment to directly refer to another
experiment results, thus a dependency chain can be formed to ensure
execution order.
It also simplifies the dataset definition, as they can be automatically
fetched from the experiment directly.