Continual Reinforcement Learning on Atari-20 benchmark leaderboard · SOTA2 Research

Continual Reinforcement Learning on Atari-20

42.7FT GainAMC

Evaluation Results

Method	Links
AMC2026.04		42.7	7.5	22	—	—
AMC2026.04		22.4	—	-9.7	201	80
PackNet2026.04		15.7	—	-12.4	187	70
PER2026.04		8.3	—	-38.6	142	45
EWC2026.04		5.2	—	-41.2	136	40
VR2026.04		2.1	—	-47.3	124	35