Event
SenseTime Astra Improves Performance, Robot Body Value Sparks Discussion
1 reports · 1 sources
SenseTime Astra achieved 100% on the LIBERO-Long++ benchmark and 81.9% on RoboDojo. Based on this, OFweek Robotics argues that the stronger a general-purpose brain becomes, the easier it is for robot bodies that conform to standard interfaces to be replaced as "peripherals," and Chinese robotics companies need to determine whether their own "body" is an asset or a peripheral. The article cites Unitree's real-machine data flywheel, AgiBot's allocation of 3/4 of R&D investment to the large and small brains, and Galaxy General's AstraBrain-WBC 0.5, trained on 2 billion frames of motion-capture data, as examples, and proposes three judgment indicators: depth of in-house development at the execution layer, data return rate, and scenario replacement cost. Galaxy General says the model's zero-shot motion tracking success rate is 92.58%, with an inference latency of 0.39 milliseconds.
Report timeline
- The stronger Astra gets, the more one question matters: Is a Chinese robot's "body" an asset or a peripheral?
SenseTime's Astra reaches 100% on LIBERO-Long++ and 81.9% on RoboDojo; the stronger the general-purpose brain, the more easily a robot body that conforms to standard interfaces can be swapped out as a "peripheral." Using Unitree's real-machine data flywheel, Zhiyuan's 3/4 R&D investment in the large and small brains, and Galaxy General's AstraBrain-WBC 0.5 trained on 2 billion frames of motion-capture data (92.58% zero-shot motion-tracking success rate, 0.39 ms inference latency) as examples, the article proposes three criteria: depth of in-house development at the execution layer, data return rate, and scenario replacement cost.
OFweek Robotics ·