Practice Makes Precise: At CoRL, NVIDIA and Partners Reveal How Robots Rehearse Delicate Work in Simulation
Practice Makes Precise: At CoRL, NVIDIA and Partners Reveal How Robots Rehearse Delicate Work in Simulation
At the Conference on Robot Learning (CoRL) in Austin, Texas, NVIDIA introduced new open robotics technologies including GR00T 2, a frontier world action model built on NVIDIA Cosmos, and Isaac Lab 3.0 with enhanced integration of the Newton physics engine. These tools aim to advance robot autonomy for delicate manipulation tasks like assembling small components and handling deformable objects. NVIDIA presented two papers among six finalists for Best Paper Awards: GRAIL for digital generation pipelines and RoboTTT for scaling visuomotor context. Partners including Ambi Robotics, Bosch Research, Siemens, and others are evaluating these technologies for industrial applications such as factory automation, robotic laundry, and warehouse operations. The release includes open workflows like Video to Data for dataset creation and the EgoHand model for 3D hand motion reconstruction.
Source: NVIDIA — Robotics · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S2
- Data origin
- Human egocentric dataOpen source S7
Reported numbers
training hours
100,000 hours
View original evidence
world action model built on NVIDIA Cosmos. Trained on diverse robot and human data, including 100,000 hours of NVIDIA open source human egocentric data , the model enables robots to learn dexterous manipulation skills. Key features include: Out-of-the-box performance: Completed 40 of 50 evaluated tasks at least once without task-specific demonstrations. Skill transfer: Learns from human demonstrations and transfers s
Open source S7GPU time reduction factor
39
View original evidence
built in Isaac Lab, demonstrates how efficient practice can translate into precision. A single policy learns all 20 tasks on the NIST Assembly Task Board, manipulating millimeter-scale parts using only a yes-or-no success signal. It achieves more than 98% training success, compared with 81% for separate per-task policies, using roughly 39x less GPU time. https://blogs.nvidia.com/wp-content/uploads/2026/11/alphanist-6
Open source S19tasks evaluated
50
View original evidence
kills between different robot bodies without robot demonstrations Generalization with little data: Achieved 71% average success across 50 tasks with just one demonstration per task, increasing to 95% with five. Agent control: AI agents can use GR00T 2’s application programming interface and supporting tools to run skills in sequence and recover from errors. https://blogs.nvidia.com/wp-content/uploads/2026/11/groot-2-
Open source S8
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
Simulation, which has long been the proving ground for robots learning to walk, roll and hop, is now helping researchers teach robots tasks that require a finer touch: aligning gears, connecting cables and assembling small components. The shift from locomotion to manipulation is taking center stage at this week’s Conference on Robot Learning (CoRL) in Austin, Texas, alongside the rise of agentic robotics — a new para
Open source S2
world action model built on NVIDIA Cosmos. Trained on diverse robot and human data, including 100,000 hours of NVIDIA open source human egocentric data , the model enables robots to learn dexterous manipulation skills. Key features include: Out-of-the-box performance: Completed 40 of 50 evaluated tasks at least once without task-specific demonstrations. Skill transfer: Learns from human demonstrations and transfers s
Open source S7
kills between different robot bodies without robot demonstrations Generalization with little data: Achieved 71% average success across 50 tasks with just one demonstration per task, increasing to 95% with five. Agent control: AI agents can use GR00T 2’s application programming interface and supporting tools to run skills in sequence and recover from errors. https://blogs.nvidia.com/wp-content/uploads/2026/11/groot-2-
Open source S8
built in Isaac Lab, demonstrates how efficient practice can translate into precision. A single policy learns all 20 tasks on the NIST Assembly Task Board, manipulating millimeter-scale parts using only a yes-or-no success signal. It achieves more than 98% training success, compared with 81% for separate per-task policies, using roughly 39x less GPU time. https://blogs.nvidia.com/wp-content/uploads/2026/11/alphanist-6
Open source S19
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Check whether the intended buyer needs human observations, hand or object annotations, or aligned robot actions. Human video alone does not provide robot commands.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：NVIDIA — Robotics · blogs.nvidia.com