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.