Offline Reinforcement Learning on D4RL MuJoCo v0 (Random, Medium-Replay, Medium, Expert Configurations) benchmark leaderboard · SOTA2 Research

Offline Reinforcement Learning on D4RL MuJoCo v0 (Random, Medium-Replay, Medium, Expert Configurations)

35.4HalfCheetah Return (Random)MOPO

Evaluation Results

Method	Links
MOPO2022.06		35.4	11.7	13.6	53.1	67.5	39	42.3	28	17.8	63.3	23.7	44.6	—	—	—	—
FisherBRC2022.06		32.2	11.4	0.6	43.3	35.6	42.6	41.3	99.4	79.5	96.1	90.6	103.6	106.8	112.3	79.9	974.6
CQL2022.06		21.7	10.7	2.7	41.9	28.6	15.8	37.2	44.2	57.5	27.1	111.4	68.1	82.4	111.2	103.8	764.3
CABI+IQL2022.06		18.4	11.4	8	42.2	36.8	17.2	41.6	40	55.1	96.7	112.8	104.8	104.7	112.8	106.8	909.3
IQL2022.06		16.2	9.3	6.2	40.5	33.4	15.8	41.2	30.7	50.8	89	111.5	99.7	100.8	112	103.6	860.7
UWAC2022.06		2.3	9.8	3.8	38.9	18	8.4	37.4	30.3	17.4	40.6	95.4	14.8	104	109.1	88.4	618.6
BCQ2022.06		2.2	10.6	4.9	38.2	33.1	15	40.7	54.5	53.1	64.7	110.9	57.5	89.9	107	102.3	784.6
BC2022.06		2	9.5	1.2	34.7	19.7	8.3	36.6	30	11.4	67.6	89.6	12	105.2	111.5	56	595.3