ResearchBenchmarksReinforcement Learning on Atari Space InvadersFollow54,500Mean Episode ReturnR2D2-1,368.813,135.627,64042,144.4Feb 14, 2020Nov 1, 2020Jul 20, 2021Apr 7, 2022Dec 24, 2022Sep 11, 2023May 29, 2024Evaluation ResultsMethodMethodLinksMean Episode ReturnR2D2Extension=RNDExtension=RND2020.0254,500NGUN=1, Variant=without RNDN=1, Variant=without RND2020.0245,300NGUN=32N=322020.0244,600R2D22020.0243,200NGUN=1N=12020.0243,000R2D2Loss=RetraceLoss=Retrace2020.0235,000Human2020.021,600Average Human2024.051,000DDQN2024.05800FDQN2024.05795DQN2024.05780