Offline Goal-conditioned Reinforcement Learning on OGBench (test) benchmark leaderboard · SOTA2 Research

Offline Goal-conditioned Reinforcement Learning on OGBench (test)

72.8PointMaze Giant NavigateMQE

Evaluation Results

Method	Links
MQE2025.11		72.8	59.2	67.7	35.1	48.6	27.6	46.5	26.5	40.8	18.3	18.7	76.8	30.8	38.1	19.8	25.9	25	17.9	26.9	36.3
QRL2025.11		68.5	49.7	0	0.4	0	0	0.4	0.4	0.4	3.4	0.4	5.4	9.1	5.4	16.3	6.1	8.6	0	0	9.8
HIQL2025.11		45.9	0	3.9	1.8	0	0	12.5	3.3	6.4	2.6	7.4	38.2	25.2	49.9	21	59.4	21	60.1	0.2	18.7
CMD2025.11		45.3	8.1	0.8	2	22.5	0.2	5	0.2	0.2	1.5	0.2	1.2	4	16.1	18.9	0.3	16.1	7.2	22.3	8.7
TMD2025.11		39.9	9.1	0.9	2.7	22.3	0	9.2	6.3	13.1	2.1	10	21.9	19.6	20.7	17.9	14.2	17.7	9.8	14.5	13
CRL2025.11		27.4	0	0.3	0	14.6	0	0.7	1.5	1.5	2.7	1.5	18.6	1.2	9.6	16.9	6	15.6	9.6	0.1	6.2
GCIQL2025.11		0	0	0.4	0	0	0	0.5	1.5	40.2	1.8	25.7	51.3	25.9	12.2	15.2	21.6	12.5	16.2	0	11.8
n-SAC+BC2025.11		0	0.4	0.2	9.2	0.3	0.5	3.2	1.7	19.1	1.4	11.4	17.6	19.1	7.1	21.1	22.7	17.1	10.3	7.6	8.5