Meta-World

Benchmarks

Task Name	Dataset Name	SOTA Result
Open door	Meta-World	VOC Score62.64	35
Reward Modeling	Meta-World Open door	Prediction Accuracy65.46	28
Reward Modeling	Meta-World Open drawer	Prediction Accuracy69.01	28
Reward Modeling	Meta-World Button press	Prediction Accuracy76.44	28
Open drawer	Meta-World	VOC Score90.17	28
Button press	Meta-World	VOC Score95.47	28
Robotic Manipulation	Meta-World	Success Rate (Easy)89.29	27
Robotic Manipulation	Meta-World	Average Success Rate94.7	27
Door Open	Meta-World	Door Open Success Rate100	20
window-open	Meta-World window-open	ASR90	20
window-close	Meta-World window-close	ASR100	20
Multi-task Reinforcement Learning	Meta-World MT50 V2	Overall Success Rate90.9	16
Multi-task Reinforcement Learning	Meta-World MT10 V2	Success Rate99.5	15
Robot Manipulation	Meta-World	Latency (Easy) (ms)10.1	15
Door Unlock	Meta-World	Success Rate96	14
Handle Press	Meta-World	Success Rate100	14
Door Lock	Meta-World	Success Rate98	14
Robotic Manipulation	Meta-World v2	Success Rate96.9	14
push	Meta-World ML-1 (test)	Success Rate1	12
Multi-task Reinforcement Learning	Meta-World MT10 v1 (Fixed)	Success Rate88	12
Multi-Task Reinforcement Learning	Meta-World MT50 V1 (final-checkpoint)	Success Rate (IQM)79.3	11
Continual Reinforcement Learning	Meta-World MT50 v2	AP81.7	11
Offline Reinforcement Learning	Meta-World medium-replay	BP->DC*3,967	10
Robot Manipulation	Meta-World	Button Success Rate100	9
Embodied AI	Meta-World 48 tasks	Success Rate70.9	9

Showing 25 of 134 rows