Benchmark
RoboDojo
Can a robot manipulation policy remember earlier observations, handle unfamiliar objects and finish a sequence of precise movements?
Explore the original evidence
Explore the original evaluation, its methods, and reported findings.
Open evalCoverage
RoboDojo includes 42 simulated tasks and 18 physical tasks. The project is operated by AI MMLab Club. XPolicyLab connects policies to both settings.
Methodology
The simulation score gives equal weight to each of five capabilities. On real robots, three evaluators score both intermediate progress and completion across repeated trials. The physical and simulated tasks are different sets. Their results cannot measure transfer on matched tasks.