Physical AI company RLWRLD (RLWRLD) said on June 8 it will develop next-generation industry standards for humanoid robot AI with Nvidia.

RLWRLD said it will strengthen cooperation along three pillars: DexBench, a benchmark for evaluating dexterity performance; setting data standards for learning precision manipulation; and integration with the open NVIDIA Isaac Lab and Isaac Lab-Arena frameworks.

The company said dexterity AI, which enables precise robot hand movements for delicate tasks such as assembly, sorting and packaging, is emerging as a key challenge in humanoid AI development. But the industry lacks common evaluation standards to objectively measure and compare dexterity performance, as well as shared data standards for large-scale training. As a result, neither technological progress nor commercialisation has been able to gain speed.

RLWRLD said the goal of the collaboration is to address both gaps at once by combining RLWRLD's evaluation methodologies and data infrastructure with Nvidia's Isaac simulation platform to establish open standards that can be adopted across the industry.

RLWRLD Chief Executive Ryu Jung-hee (류중희) said, "Without a common language to measure and reproduce robot hand movements, there are limits to the commercialisation of dexterity A." He added, "By establishing DexBench and data standards with Nvidia, RLWRLD has taken on a role in designing infrastructure for the entire industry beyond simply developing models. This cooperation will become a new benchmark for the global humanoid AI ecosystem."