I was pretty sure it is hard but in starter kit sources I saw:
“distribution_mode” : “easy” # What variant of the levels to use, the options are “easy”, “hard”, “extreme”, “memory”, “exploration”. All games support “easy” and “hard”, while other options are game-specific. The default is “hard”. Switching to “easy” will reduce the number of timesteps required to solve each game and is useful for testing or when working with limited compute resources. NOTE : During the evaluation phase (rollout), this will always be overriden to “easy”
Hi @denys88,
I can confirm that we are using "distribution_mode": "easy" during the rollout/evaluation phase as mentioned.
Assumption: In case you are not adding distribution_mode in the training phase and judging based on the training stage value, it may be running as hard (i.e. default value).
Can you share observation or logs based on which you think it is configured to hard?
Happy to debug and verify it again on our side! 
Thank you very much for the answer.
I misunderstood that: ’ In both cases, agents will be evaluated on the full distribution of levels.’