- DexBench, 5-Finger Dexterity Humanoid Data Standards, and integration of NVIDIA Isaac libraries — Advancing the Open Dexterity AI Ecosystem
- No Measurement Framework, No Data Standard — RLWRLD Sets Out to Build the Industry's Common Foundation with NVIDIA
SAN FRANCISCO, June 9, 2026 /PRNewswire/ -- RLWRLD (CEO: Junghee Ryu), a physical AI company, announced today a collaboration with NVIDIA to develop next-generation industry standards for humanoid robot AI. RLWRLD will focus on three pillars: DexBench, a universal benchmark for evaluating dexterity performance; data standard for dexterous manipulation training; and deep integration with the open NVIDIA Isaac Lab and Isaac Lab-Arena frameworks.
Dexterous manipulation — enabling humanoid robots to perform fine-grained tasks such as precision assembly, sorting, and packaging — has emerged as the decisive frontier in humanoid AI development. Yet the industry lacks both a common framework for objectively measuring and comparing humanoid dexterity performance, and a shared data standard for training dexterous manipulation models at scale — gaps that slow both technology development and commercial deployment.
Junghee Ryu, CEO of RLWRLD, said: "Without a shared language for measuring and reproducing the precise movements of a robot hand, the commercial potential of dexterity AI remains constrained. By establishing DexBench and a data standard with NVIDIA, RLWRLD is stepping beyond model development to architect the infrastructure of an entire industry. We are confident this collaboration will set a new reference point for the global humanoid AI ecosystem."
Amit Goel, Head of Robotics Ecosystem at NVIDIA said: "Measurable and reproducible dexterous manipulation is essential to scaling robotics adoption in industrial environments. DexBench's integration of the NVIDIA Isaac platform provides the robotics community with the standardized metrics and data infrastructure required to accelerate the development of reliable, high-precision manipulation."
DexBench Development… Dual-Validation Framework Spanning Simulation and Real-World Settings
RLWRLD's DexBench benchmark will be integrated into NVIDIA's Isaac Lab-Arena environment, establishing a system for validating dexterity performance across both simulation and real-world conditions. DexBench was developed directly from dexterous manipulation tasks observed in industrial environments, and defines five core evaluation domains — Grasp Diversity, Spatial Precision, Temporal Precision, Contact Precision, and Context Awareness — spanning 18 Key Atomic Tasks. Standardized evaluation metrics grounded in actual industrial tasks including assembly, sorting, and packaging will be proposed as an open industry specification — giving robot manufacturers, researchers, and enterprises a common yardstick for the first time, and a clear path from benchmark performance to commercial deployment.
5-Finger Dexterity Humanoid Data Standard… Building a Shared Foundation for Global Robot Manufacturers and Research Institutions
A data format for dexterous manipulation training will be defined in collaboration with NVIDIA to ensure compatibility with NVIDIA Isaac Lab pipelines. The standard aims to serve as a common data interface for global robot manufacturers and research organizations worldwide.
RLDX-1, RLWRLD's foundation model for humanoid dexterous manipulation, has demonstrated state-of-the-art performance across 8 established simulation benchmarks — including RoboCasa Kitchen, RoboCasa GR-1 Tabletop, and LIBERO-Plus — outperforming frontier models such as NVIDIA GR00T N1.6 and Physical Intelligence π₀.₅. These results validate RLWRLD's architectural approach, while DexBench addresses the next frontier: standardizing how the industry measures dexterity performance that existing benchmarks were not designed to capture.
RLWRLD has been expanding its global footprint through a series of RLDX-1 launch events under the "Dexterity Night" banner. At the San Francisco debut last month, NVIDIA's Head of Robotics Ecosystem and Edge AI Product Amit Goel took the stage and called RLWRLD "one of the core partners in the physical AI ecosystem we are building at NVIDIA," drawing strong interest from the global robotics community. Following subsequent events in Japan, the company will next bring Dexterity Night to Seoul on June 10.
About RLWRLD
RLWRLD is a physical AI company developing robotics foundation models that bring human-level dexterity and cognition to machines. Founded in 2024 and headquartered in the United States, with additional offices in Korea and Japan, RLWRLD operates a proprietary system for collecting and training on high-precision multimodal industrial data, enabling robots to perceive, understand and act in real-world industrial environments. The company is engaged in commercialization pilots with leading industry partners and aims to become the global leader in industrial robotics AI.
SAN FRANCISCO, June 9, 2026 /PRNewswire/ -- RLWRLD (CEO: Junghee Ryu), a physical AI company, announced today a collaboration with NVIDIA to develop next-generation industry standards for humanoid robot AI. RLWRLD will focus on three pillars: DexBench, a universal benchmark for evaluating dexterity performance; data standard for dexterous manipulation training; and deep integration with the open NVIDIA Isaac Lab and Isaac Lab-Arena frameworks.
Dexterous manipulation — enabling humanoid robots to perform fine-grained tasks such as precision assembly, sorting, and packaging — has emerged as the decisive frontier in humanoid AI development. Yet the industry lacks both a common framework for objectively measuring and comparing humanoid dexterity performance, and a shared data standard for training dexterous manipulation models at scale — gaps that slow both technology development and commercial deployment.
Junghee Ryu, CEO of RLWRLD, said: "Without a shared language for measuring and reproducing the precise movements of a robot hand, the commercial potential of dexterity AI remains constrained. By establishing DexBench and a data standard with NVIDIA, RLWRLD is stepping beyond model development to architect the infrastructure of an entire industry. We are confident this collaboration will set a new reference point for the global humanoid AI ecosystem."
Amit Goel, Head of Robotics Ecosystem at NVIDIA said: "Measurable and reproducible dexterous manipulation is essential to scaling robotics adoption in industrial environments. DexBench's integration of the NVIDIA Isaac platform provides the robotics community with the standardized metrics and data infrastructure required to accelerate the development of reliable, high-precision manipulation."
DexBench Development… Dual-Validation Framework Spanning Simulation and Real-World Settings
RLWRLD's DexBench benchmark will be integrated into NVIDIA's Isaac Lab-Arena environment, establishing a system for validating dexterity performance across both simulation and real-world conditions. DexBench was developed directly from dexterous manipulation tasks observed in industrial environments, and defines five core evaluation domains — Grasp Diversity, Spatial Precision, Temporal Precision, Contact Precision, and Context Awareness — spanning 18 Key Atomic Tasks. Standardized evaluation metrics grounded in actual industrial tasks including assembly, sorting, and packaging will be proposed as an open industry specification — giving robot manufacturers, researchers, and enterprises a common yardstick for the first time, and a clear path from benchmark performance to commercial deployment.
5-Finger Dexterity Humanoid Data Standard… Building a Shared Foundation for Global Robot Manufacturers and Research Institutions
A data format for dexterous manipulation training will be defined in collaboration with NVIDIA to ensure compatibility with NVIDIA Isaac Lab pipelines. The standard aims to serve as a common data interface for global robot manufacturers and research organizations worldwide.
RLDX-1, RLWRLD's foundation model for humanoid dexterous manipulation, has demonstrated state-of-the-art performance across 8 established simulation benchmarks — including RoboCasa Kitchen, RoboCasa GR-1 Tabletop, and LIBERO-Plus — outperforming frontier models such as NVIDIA GR00T N1.6 and Physical Intelligence π₀.₅. These results validate RLWRLD's architectural approach, while DexBench addresses the next frontier: standardizing how the industry measures dexterity performance that existing benchmarks were not designed to capture.
RLWRLD has been expanding its global footprint through a series of RLDX-1 launch events under the "Dexterity Night" banner. At the San Francisco debut last month, NVIDIA's Head of Robotics Ecosystem and Edge AI Product Amit Goel took the stage and called RLWRLD "one of the core partners in the physical AI ecosystem we are building at NVIDIA," drawing strong interest from the global robotics community. Following subsequent events in Japan, the company will next bring Dexterity Night to Seoul on June 10.
About RLWRLD
RLWRLD is a physical AI company developing robotics foundation models that bring human-level dexterity and cognition to machines. Founded in 2024 and headquartered in the United States, with additional offices in Korea and Japan, RLWRLD operates a proprietary system for collecting and training on high-precision multimodal industrial data, enabling robots to perceive, understand and act in real-world industrial environments. The company is engaged in commercialization pilots with leading industry partners and aims to become the global leader in industrial robotics AI.
** This press release is distributed by PR Newswire through automated distribution system, for which the client assumes full responsibility. **
RLWRLD Launches DexBench Initiative to Define Next-Generation Industry Standards for Humanoid AI in Collaboration with NVIDIA
