Introduction of Helix 2.5: Achieving Zero-Shot Generalization in Humanoid Robotics
- Published
- Source
- figure.ai
Figure has unveiled Helix 2.5, a groundbreaking neural network designed for humanoid robots. Unlike previous models, Helix 2.5 can autonomously perform tasks in unfamiliar environments without prior data collection. This innovation was tested in 30 homes in the Bay Area, showcasing the robot's ability to execute three complex behaviors: tidying living rooms, folding towels, and making beds.
The significance of Helix 2.5 lies in its ability to generalize learned behaviors across different settings, a challenge that has historically limited robotic applications. By utilizing the Index dataset, which encompasses a wide range of human behaviors, Helix 2.5 achieved a zero-shot success rate of 56%, a substantial improvement from the previous 9%. This advancement indicates a shift towards more adaptable and intelligent robotic systems capable of operating in diverse environments.
Looking ahead, the implications of Helix 2.5's capabilities could transform the landscape of humanoid robotics. While the technology is not yet fully mature, it represents a significant step towards developing robots that can learn from human experiences and apply that knowledge in new scenarios. Future developments will likely focus on expanding the range of tasks and environments in which these robots can operate effectively.
Original report
Helix 2.5: Zero-Shot 30-Home Generalization
figure.ai · Figure AIThis briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
Related Suppliers
Figure AI
United States
American humanoid robotics company developing autonomous general-purpose robots powered by vision-language-action AI models.
Shanghai Jiao Tong University Underwater Engineering Institute Co., Ltd. 上海交大海洋水下工程科学研究院有限公司
China
SJTU-owned Shanghai institute (est. 1978) behind the Haima-4500 and Hailong 3,500 m deep-sea ROVs, also offering underwater engineering and diving services.
Marine Tech SAS
France
Six-Fours-les-Plages, France marine technology firm with over 20 years' experience; builds RSV surface drones, UHV Manta hybrid vehicles, and water systems.
Mytra
United States
US warehouse robotics company based in Brisbane, CA, deploying Helix-design bots that lift 3,000 lbs in slotless, software-defined ASRS up to 80 ft tall.
Copperstone Technologies
Canada
Canadian robotics company producing autonomous amphibious robots for hazardous industrial and mining environments.
Boardwalk Robotics
United States
Pensacola-based humanoid robotics company developing Alex, a 19-DOF legless upper-torso humanoid for manufacturing and logistics.
Cyborg Robotics 赛博格机器人
China
Shenzhen humanoid robot maker focused on embodied intelligence for hazardous industry; produces full-size biomimetic humanoids and dexterous hands.
Jiying Technology 芨影科技
China
China-based AI company known for Jiying 2.0, a zero-shot generalizable physics foundation model launched in July 2026 for engineering simulation.
Physical Intelligence
United States
San Francisco AI robotics company founded 2024; its pi0 vision-language-action models direct many robot types, backed by billions in venture funding.
Encord
United Kingdom
London-based AI data platform for physical AI teams — annotate, curate, and evaluate multimodal datasets for robots and autonomous systems.
Pix4D Pix4D
Switzerland
Swiss photogrammetry software spun out of EPFL 2011; converts drone imagery into 2D maps and 3D models for survey, agriculture, and construction.
Flexion Robotics
Switzerland
Swiss startup building a hardware-agnostic AI autonomy stack for humanoid robots using simulation and reinforcement learning.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.
