Wuwen Xinqiong, together with Tsinghua and AgiBot, proposes Collective Experience Learning (CEL), with a 98.3% success rate verified on 96 robots.
Wuwen Xinqiong, together with Tsinghua University and AgiBot, proposes the Collective Experience Learning (CEL) paradigm and completes prototype validation on 96 robots. In a phone packaging task, the policy trained with CEL completed 120 tests on two same-model robots that did not participate in training, with the success rate rising from 79.2% to 98.3%; when the number of robots participating in learning increased from 16 to 96, the test success rate rose from 86.7% to 98.3%.
Highlights
Qiyuan releases the PrimeSim world model simulator, topping EWMBench and WorldSimProbe.
The Qiyuan Embodied Research Center releases PrimeSim, a world model simulator for robot manipulation tasks, built on a pretrained video generation model through two-stage mid-training and post-training adaptation. It ranks first on EWMBench with a composite score of 6.09, and takes first place in three of five WorldSimProbe evaluations: global trajectory coverage, action-source behavior preservation, and interaction dynamics.
PokeBot releases the embodied API RoboAPI, and the world-action model Poke-Aurora makes its debut.
PokeBot released its first model product RoboAPI in the early hours of October 30, a publicly accessible, out-of-the-box embodied API. After developers connect and adapt a robotic arm, they need neither to collect data nor to train models; a single prompt command is enough to execute tasks that were not specifically trained for.
AgiBot releases the WITA-OMNI interaction model and Lingqu OS 1.0, to be fully open-sourced by year-end.
At the 2026 Innovation Day, AgiBot released the embodied omni-modal end-to-end interaction large model WITA-OMNI, which ranks first in four evaluations: Daily-Omni, TimeLens-Bench, EmbSpatialBench, and MMAU, with an average response of about 600 milliseconds.
机器人前瞻 · 微信公众号Source ↗
Model releases/updates
Qiyuan releases the PrimeSim world model simulator, topping EWMBench and WorldSimProbe.
The Qiyuan Embodied Research Center releases PrimeSim, a world model simulator for robot manipulation tasks, built on a pretrained video generation model through two-stage mid-training and post-training adaptation. It ranks first on EWMBench with a composite score of 6.09, and takes first place in three of five WorldSimProbe evaluations: global trajectory coverage, action-source behavior preservation, and interaction dynamics.
PokeBot releases the embodied API RoboAPI, and the world-action model Poke-Aurora makes its debut.
PokeBot released its first model product RoboAPI in the early hours of October 30, a publicly accessible, out-of-the-box embodied API. After developers connect and adapt a robotic arm, they need neither to collect data nor to train models; a single prompt command is enough to execute tasks that were not specifically trained for.
AgiBot releases the WITA-OMNI interaction model and Lingqu OS 1.0, to be fully open-sourced by year-end.
At the 2026 Innovation Day, AgiBot released the embodied omni-modal end-to-end interaction large model WITA-OMNI, which ranks first in four evaluations: Daily-Omni, TimeLens-Bench, EmbSpatialBench, and MMAU, with an average response of about 600 milliseconds.
机器人前瞻 · 微信公众号Source ↗
Product releases/updates
Critical Point releases the OmniGlove Pro data-collection glove, and the OmniHand dexterous hand completes production-line validation across 72 processes.
At the dexterous manipulation sub-forum of AgiBot Innovation Day, Critical Point released the OmniGlove Pro data-collection glove and began mass-production delivery, proposing the "real-machine Ready" standard for data qualification, whereby a robot can reproduce a task in the same scenario. In the same period, the OmniHand dexterous hand continuously completed 72 processes at a POC workstation on the Shaoxing Wolong production line, with 8.5 minutes per round, a 98% task success rate, and the ability to work continuously for 8 hours.
36Kr Embodied IntelligenceSource ↗
Industry dynamics
AgiBot Innovation Day releases the Worken work-hour metric and Lingqu OS 1.0.
At AgiBot Innovation Day on October 29, AgiBot proposed the Worken (effective intelligent work-hour) metric, with the formula: number of active robots × autonomous working hours × output efficiency ratio, and released the embodied intelligence native operating system Lingqu OS 1.0.
雷科技Source ↗Agility Robotics and Churchill Capital Corp XI expect to raise approximately US$465 million and complete their merger.
Agility Robotics and Churchill Capital Corp XI announced that the expected total financing for the merger and related financing is approximately US$465 million (before deducting transaction expenses), of which approximately US$164 million comes from the Churchill XI trust account, approximately US$201 million from a PIPE at US$10 per share (led by Foxconn), and US$100 million from equity financing already received after the merger announcement.
Paper research
Wuwen Xinqiong, together with Tsinghua and AgiBot, proposes Collective Experience Learning (CEL), with a 98.3% success rate verified on 96 robots.
Wuwen Xinqiong, together with Tsinghua University and AgiBot, proposes the Collective Experience Learning (CEL) paradigm and completes prototype validation on 96 robots. In a phone packaging task, the policy trained with CEL completed 120 tests on two same-model robots that did not participate in training, with the success rate rising from 79.2% to 98.3%; when the number of robots participating in learning increased from 16 to 96, the test success rate rose from 86.7% to 98.3%.
Yunzhu Li's team at Columbia University presents progress on the open-source tactile sensor FlexiTac at IROS 2026.
Yunzhu Li's team at Columbia University presented at IROS 2026 the progress of the open-source flexible piezoresistive tactile sensor platform FlexiTac one and a half years after open-sourcing; the sensor can be assembled in 3 minutes, can be bent and cut, and can wrap around gloves, fingertips, or robot skin. The team has integrated it into LeRobot and modified the UMI handheld device to add tactile recording, for synchronized visual-tactile data collection and policy pretraining. ADI is advancing a commercial version, with tactile resolution five times that of a human finger, mass-produced based on semiconductor manufacturing processes.
Leiphone RoboticsSource ↗