Zhongke Huiling completes a round of A++ financing worth several hundred million yuan, advancing embodied intelligence mass-production delivery
Zhongke Huiling completed an A++ round of financing worth several hundred million yuan, jointly invested by Zhilai Investment, Yonghua Investment, Maotian Investment, Yuanhe Chenkun, Guoke Kezhi, Luning Investment and others, with the funds used for product mass-production delivery, embodied brain technology R&D iteration and market expansion. The company's self-developed CASBOT PA perception-to-action model has been validated at actual workstations in a shoe factory in Putian, with the outsole stamping process reaching a cycle time of 7 seconds per piece and sole pressing reaching 18 seconds per pair. In 2026, its total in-hand orders for embodied shoe-making robots exceeded 400 million yuan, and its equipment production scheduling scale entered the thousands-of-units range.
Highlights
RepresentNav: a navigation foundation model built on a self-constructed representation space
The authors propose RepresentNav, a navigation foundation model that uses a three-layer transformer of Representation Residuals to aggregate full-depth features from multiple visual foundation models, constructing a representation space driven by the policy's own objectives.
arXiv cs.ROSource ↗Dream Labs raises a $90 million seed round and launches Dream API and the Dream-1 model
Dream Labs announced the completion of a $90 million seed round, led by B Capital with participation from Kindred Ventures and others. The company also launched Dream API, which can convert robot camera images and natural-language instructions directly into motor commands, and said its flagship model Dream-1 is trained on videos of everyday human tasks and can zero-shot control any robot equipped with a gripper.
Robotics & Automation NewsSource ↗RoboCoach: using a world model to locate weak subtasks and targeted teaching to improve long-horizon task success rates
The team of Zhang Changshui and Cui Sen at Tsinghua University proposes RoboCoach, an embodied intelligence system that uses the world model CoachWorld to run long-horizon tasks in closed-loop trial runs in imagination, locate the first timed-out weak subtask and request targeted human teaching, updating only the corresponding skill expert. Under a fixed demonstration budget, the success rate rose from 13.3% to 75.0% on Franka and from 40.0% to 83.8% on AgileX, while the uniform-collection baseline was only 30.0% and 47.5% respectively. The average success rate on four unseen new task routes was 35.0%, versus 0% for the shared-policy baseline.
机器之心 · 微信公众号Source ↗
Industry News
Zhongke Huiling completes a round of A++ financing worth several hundred million yuan, advancing embodied intelligence mass-production delivery
Zhongke Huiling completed an A++ round of financing worth several hundred million yuan, jointly invested by Zhilai Investment, Yonghua Investment, Maotian Investment, Yuanhe Chenkun, Guoke Kezhi, Luning Investment and others, with the funds used for product mass-production delivery, embodied brain technology R&D iteration and market expansion. The company's self-developed CASBOT PA perception-to-action model has been validated at actual workstations in a shoe factory in Putian, with the outsole stamping process reaching a cycle time of 7 seconds per piece and sole pressing reaching 18 seconds per pair. In 2026, its total in-hand orders for embodied shoe-making robots exceeded 400 million yuan, and its equipment production scheduling scale entered the thousands-of-units range.
Dream Labs raises a $90 million seed round and launches Dream API and the Dream-1 model
Dream Labs announced the completion of a $90 million seed round, led by B Capital with participation from Kindred Ventures and others. The company also launched Dream API, which can convert robot camera images and natural-language instructions directly into motor commands, and said its flagship model Dream-1 is trained on videos of everyday human tasks and can zero-shot control any robot equipped with a gripper.
Robotics & Automation NewsSource ↗GXO and Agility Robotics expand partnership, will deploy Digit humanoid robots at North Carolina and Tennessee sites
GXO and Agility Robotics are expanding their partnership and will deploy Digit humanoid robots at sites in Statesville, North Carolina, and Lebanon, Tennessee, to handle repetitive material-handling tasks in e-commerce fulfillment, with deployment of Digit 4 expected to begin in early 2027 and a migration once Digit 5 is commercially available.
- Agility Robotics announces Digit commercial progress: GXO expands to two new facilities, adding three manufacturing customers
Paper Research
RepresentNav: a navigation foundation model built on a self-constructed representation space
The authors propose RepresentNav, a navigation foundation model that uses a three-layer transformer of Representation Residuals to aggregate full-depth features from multiple visual foundation models, constructing a representation space driven by the policy's own objectives.
arXiv cs.ROSource ↗RoboCoach: using a world model to locate weak subtasks and targeted teaching to improve long-horizon task success rates
The team of Zhang Changshui and Cui Sen at Tsinghua University proposes RoboCoach, an embodied intelligence system that uses the world model CoachWorld to run long-horizon tasks in closed-loop trial runs in imagination, locate the first timed-out weak subtask and request targeted human teaching, updating only the corresponding skill expert. Under a fixed demonstration budget, the success rate rose from 13.3% to 75.0% on Franka and from 40.0% to 83.8% on AgileX, while the uniform-collection baseline was only 30.0% and 47.5% respectively. The average success rate on four unseen new task routes was 35.0%, versus 0% for the shared-policy baseline.
机器之心 · 微信公众号Source ↗Study: VLA full-weight fine-tuning without an fp32 master copy silently erases most updates
The paper points out that when VLA models are full-weight fine-tuned in bf16 without retaining an fp32 master copy, AdamW directly rounds off updates smaller than 1/512 of the weight; 9 of 12 open-source VLA codebases train this way by default or in documented recipes, including LeRobot's SmolVLA default configuration and the π0/π0.5 guide.
arXiv cs.ROSource ↗