Reinforcement Learning on Atari 1M (26 Game Scores and Summary Statistics)
7,127.7Alien ScoreHuman
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Human2026.01 | 7,127.7 | 1,719.5 | 742 | 8,503.3 | 753.1 | 37,187.5 | 12.1 | 30.5 | 7,387.8 | 35,829.4 | 1,971 | 29.6 | 4,334.7 | 2,412.5 | 30,826.4 | 302.8 | 3,035 | 2,665.5 | 22,736.3 | 6,951.6 | 14.6 | 69,571.3 | 13,455 | 7,845 | 42,054.7 | 11,693.2 | — | 1 | 1 | |
| DreamerV3Training interactions=1M2026.01 | 1,978.2 | 265.5 | 1,949.1 | 2,917.3 | 569.5 | 5,261.8 | 96.6 | 25.1 | 4,442.8 | 108,468.5 | 644.4 | 7.9 | 3,741.9 | 14,935.7 | 13,600.5 | 465.6 | 5,432.7 | 13,389.3 | 35,395.8 | 3,165.6 | 9.4 | 882.4 | 5,006.7 | 21,404 | 1,580.6 | 115,045.3 | 10 | 2.213 | 0.782 | |
| EADreamTraining interactions=1M2026.01 | 1,550.2 | 672.8 | 3,369.2 | 6,605.5 | 1,272.5 | 39,070 | 93.6 | 416.8 | 647 | 112,524 | 2,323.8 | 34 | 3,989.1 | 83,756.6 | 20,564.1 | 688.5 | 5,079 | 11,534 | 46,526 | 3,449.9 | 21 | 5,595 | 14,411 | 19,359 | 2,387.8 | 416,219 | 17 | 5.273 | 1.188 | |
| EADreamTraining interactions=100K2026.01 | 776.4 | 144.2 | 883.4 | 1,096.9 | 742.6 | 13,372 | 85.4 | 71.8 | 904 | 89,038.6 | 152.2 | 0 | 692.6 | 4,415.8 | 8,801.8 | 337.2 | 3,875.6 | 8,729.6 | 23,434.6 | 1,580.7 | 20.1 | -472.5 | 1,664.4 | 12,518.6 | 557.9 | 28,408.2 | 12 | 1.29 | 0.651 | |
| Random2026.01 | 227.8 | 5.8 | 222.4 | 210 | 14.2 | 2,360 | 0.1 | 1.7 | 811 | 10,780.5 | 152.1 | 0 | 65.2 | 257.6 | 1,027 | 29 | 52 | 1,598 | 258.5 | 307.3 | -20.7 | 24.9 | 163.9 | 11.5 | 68.4 | 533.4 | 0 | 0 | 0 |