Offline Reinforcement Learning on D4RL v2 benchmark leaderboard · SOTA2 Research

Offline Reinforcement Learning on D4RL v2

60Score (HalfCheetah-M)Diffusion-DICE

Evaluation Results

Method	Links
Diffusion-DICE2024.07		60	—	—	100.2	89.3	49.2	102.3	90.8	97.3	112.2	114.1	98.1	82	91.3	85.7	68.6	72	78.3	67.8	84.4	83.8
QGPO2024.07		54.1	—	—	98	86	47.6	96.9	84.4	93.5	108	110.7	96.4	74.4	83.6	83.8	66.6	64.8	—	—	73.9	54.2
Diffusion-QL2024.07		51.1	—	—	90.5	87	47.8	101.3	95.5	96.8	111.1	110.1	93.4	66.2	76.6	78.6	46.4	56.6	60.5	62.6	72.8	57.3
IDQL2024.07		51	—	—	65.4	82.5	45.9	92.1	85.1	95.9	108.6	112.7	94	80.2	84.5	84.8	63.5	67.9	—	—	—	—
SQL2024.07		48.3	—	—	75.5	84.2	44.8	99.7	81.2	94	111.8	110	92.2	74	80.2	79.1	53.2	52.3	—	—	—	—
IQL2024.07		47.4	—	—	66.3	72.5	44.2	95.2	76.1	86.7	101.5	110.6	85.5	66.7	72.2	71	39.6	47.5	46.3	51	71.5	37.3
O-DICE2024.07		47.4	—	—	86.1	84.9	44	99.9	83.6	93.2	110.8	110.8	94.1	79.5	86	82.7	55.9	54	—	—	—	—
SfBC2024.07		45.9	—	—	57.1	77.9	37.1	86.2	65.1	92.6	108.6	109.8	92	85.3	81.3	82	59.3	45.5	—	—	—	—
CQL2024.07		44	—	—	58.5	72.5	45.5	95	77.2	90.7	105.4	109.6	84.8	43.4	65.2	54	38.4	31.6	49.8	51	37.5	39.2
DT2021.06		—	672.6	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—
TD3+BC2021.06		—	677.4	1,013.2	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—