SEOUL, South Korea -- NdotLight (CEO: Jay Park), a SimReady (Simulation-Ready) data generation solution provider for Physical AI, is joining RLWRLD in the DexBench project, a global benchmark for evaluating humanoid dexterity being co-developed with NVIDIA. NdotLight will lead the SimReady data generation to construct the simulation environment at the core of the project, building the next generation of humanoid evaluation standards together with RLWRLD.

DexBench is a next-generation, open-source benchmark for humanoid manipulation, built on NVIDIA Isaac Lab-Arena. It is designed to quantitatively measure how well robots grasp objects, assemble parts and use tools with human-like dexterity across a range of industrial settings, and high-fidelity SimReady data reflects real-world conditions is central to how well the benchmark performs and how much it can be trusted.

For this project, NdotLight will apply TRINIX, its SimReady generation solution, to build the high-precision SimReady assets DexBench requires. TRINIX automatically generates joint structures, kinematics, collision geometry and physical properties directly from CAD data, producing simulation environments that closely match real robots.

The resulting SimReady assets will feed into training and evaluation for RLDX-1, the robotics foundation model (RFM) developed by RLWRLD. The two companies plan to continuously test and narrow the sim-to-real gap, the difference between simulated and real-world robot behavior, as they work toward a more reliable, global standard for humanoid evaluation.

Korean robotics and simulation companies including Physics Sim Lab and Space AI are also contributing their expertise to the project. Among the participants, NdotLight serves as the core technology partner responsible for SimReady data generation and simulation environment construction.

"In the era of Physical AI, robot performance is determined not just by the AI model, but by the quality of the SimReady data used to train and evaluate it," said Jay Park, CEO of NdotLight. "It means a great deal that NdotLight's SimReady technology will serve as a core data technology in this global humanoid evaluation environment, through the DexBench project we are co-developing with NVIDIA. We will keep working with NVIDIA and our other global partners to build the standard data ecosystem for the Physical AI era."