The agents will be evaluated on a “secret” test set, of different env dimensions and number of agents. This is done to ensure that the solutions generalize well. We will however, release a set of generated “Levels” that are similar to the once used for evalution.
This will happen soon so stay tuned.