NR
The brief
- •GR00T 2 completed 40 of 50 tasks at least once without task-specific demonstrations; success averaged 71% with one demonstration per task and 95% with five.
- •Isaac Lab 3.0 with Newton delivered up to 2.5 times the training throughput of Isaac Lab 2.3 with PhysX in a vision-based dexterity benchmark.
- •AlphaNIST trained one policy for 20 precision-assembly tasks, achieving more than 98% training success with roughly 39 times less GPU time than separate task policies.
- •NVIDIA also introduced open evaluation benchmarks DexBench and VABAR for testing robot dexterity and performance across industrial tasks.
By Sasa Docca
