HyperAI

Atari Games On Atari 57

Metrics

Human World Record Breakthrough
Mean Human Normalized Score

Results

Performance results of various models on this benchmark

Comparison Table
Model NameHuman World Record BreakthroughMean Human Normalized Score
off-policy-actor-critic-with-shared-171741.36%
gdi-rethinking-what-makes-reinforcement229620.98%
recurrent-experience-replay-in-distributed153374.31%
learnable-behavior-control-breaking-atari2410077.52%
impala-scalable-distributed-deep-rl-with3957.34%
gdi-rethinking-what-makes-reinforcement--
generalized-data-distribution-iteration229620.33%
generalized-data-distribution-iteration177810.1%
rainbow-combining-improvements-in-deep4873.97%
mastering-atari-go-chess-and-shogi-by194996.20%
munchausen-reinforcement-learning-504%