Lightwheel unveils the RoboFinals simulation evaluation platform
Lightwheel announced RoboFinals and the RoboFinals-100 benchmark for evaluating frontier VLA and robot foundation models at scale.
Evidence notes
- RoboFinals-100 spans 100 household, factory and retail tasks with unified success criteria.
- The benchmark supports tabletop arms, mobile manipulators and full loco-manipulation systems.
- Teams can run cloud or on-premises evaluation across Isaac Lab, Newton, PhysX, MuJoCo and Genesis backends.
Company context
Lightwheel builds Physical AI infrastructure for robotics teams, frontier model developers, and industrial customers. Its platform spans SimReady assets, synthetic and egocentric human data, simulation environments, robotics evaluation, and enterprise deployment tooling.