RoboSignal Daily · 2026-10-26 · Monday
ROBOSIGNAL
HandsOn: A Benchmark for Dexterous Manipulation with Cross-Embodiment and Task Generalization
A research team from Tsinghua University, in collaboration with other institutions, has developed the HandsOn benchmark, a comprehensive evaluation framework for dexterous robotic hands. The benchmark includes 66 tasks and 18,000 human-teleoperation demonstrations, focusing on tool use, fine contact, bimanual coordination, and task transfer. It evaluates the performance of various models, including diffusion policies and visual-language-action models, and highlights the challenges in achieving stable execution and generalization across different tasks and objects.
Research
HandsOn: A Benchmark for Dexterous Manipulation with Cross-Embodiment and Task Generalization
A research team from Tsinghua University, in collaboration with other institutions, has developed the HandsOn benchmark, a comprehensive evaluation framework for dexterous robotic hands. The benchmark includes 66 tasks and 18,000 human-teleoperation demonstrations, focusing on tool use, fine contact, bimanual coordination, and task transfer. It evaluates the performance of various models, including diffusion policies and visual-language-action models, and highlights the challenges in achieving stable execution and generalization across different tasks and objects.
HandsOn: A Benchmark for Dexterous Hand Manipulation in Robotics
A research team from Tsinghua University, in collaboration with other institutions, has developed HandsOn, a comprehensive benchmark for evaluating dexterous hand manipulation in robotics. The benchmark includes 66 tasks and 18,000 human-teleoperation demonstrations, focusing on tool use, fine contact, bimanual coordination, and task transfer. The study evaluates various approaches, including diffusion policies and visual-language-action models, and reports success rates of 29.1% on Core-50 tasks and below 1% on Transfer-16 tasks. The benchmark aims to address the limitations of existing benchmarks by providing a structured evaluation of both execution and generalization capabilities.
HandsOn: A Benchmark for Dexterous Manipulation with Cross-Embodiment Evaluation
A research team from Tsinghua University, in collaboration with other institutions, has developed the HandsOn benchmark, a comprehensive evaluation framework for dexterous robotic hands. The benchmark includes 66 tasks and 18,000 human-teleoperation demonstrations, designed to assess both execution and generalization capabilities. It highlights the challenges of contact management, tool use, and multi-step coordination in robotic manipulation.
RAX: Scaling Online Exploration for Robot Agents
Researchers from Harvard, NVIDIA, UC Berkeley, MIT, and Stanford introduce RAX, a robot learning framework that enables agents to autonomously explore and accumulate skills through online exploration. RAX achieves a 37.4% average improvement over previous state-of-the-art models across five benchmarks, and demonstrates zero-shot performance on real robots without human demonstrations. The system uses code-as-policy to convert exploration into reusable skills, enabling cross-embodiment learning and deployment.
Robot Agent eXploration (RAX): Scaling Online Exploration for Self-Improving Robots
Researchers from Harvard, NVIDIA, UC Berkeley, MIT, and Stanford introduce RAX, a robot learning framework that enables continuous self-improvement through online exploration. RAX allows robots to acquire and reuse skills without human demonstrations, achieving significant performance gains across multiple benchmarks and real-world tasks. The system uses code-as-policy and skill accumulation, enabling robots to adapt to new environments and tasks without retraining or fine-tuning. RAX demonstrates average improvements of 37.4 percentage points over existing state-of-the-art models, with success rates reaching 98.5% on some tasks.
The o1 Moment in Robotics Arrives: Harvard, Berkeley, NVIDIA Introduce RAX, Enabling Cross-Embodiment Online Exploration and Skill Accumulation
A research team from Harvard University, NVIDIA, University of California, Berkeley, Massachusetts Institute of Technology, and Stanford University has introduced RAX (Robot Agent eXploration), a framework that enables robots to learn through online exploration, achieving significant performance improvements across multiple benchmarks without requiring human demonstrations. RAX achieves an average lead of 37.4 percentage points across five simulation benchmarks and can perform tasks such as precise hole insertion and manipulation of flexible objects on real robots without human demonstration. This study marks the first systematic investigation into cross-simulation, cross-robot, and cross-embodiment online exploration scaling, providing an extensible platform for the continued development of embodied intelligence.
Shanghai Hosts Real-World Testing for Embodied AI: 800 Teams Blind Test Robots in Factory Scenarios
Over 1,000 robots and 800+ teams from across China participate in a three-day 'black box blind test' at the Shanghai Zhangjiang Science Hall, focusing on real-world manufacturing, logistics, and inspection tasks. The event, part of the 2026 Global Developers' Pioneer Conference, aims to bridge the gap between lab research and industrial deployment by using real production scenarios and standardized skill modules.
Introducing the FarView A3 Humanoid Robot with Synchronized Dance Performance
Zhuyuan Robotics has released the FarView A3 humanoid robot, which successfully set a Guinness World Record for the most humanoid robots dancing simultaneously. The 110-unit deployment demonstrated synchronized movement, advanced coordination, and real-world adaptability, marking progress toward large-scale deployment in public services and entertainment.
LDB Embodied Intelligence General-Purpose Robot Released, Driving Industrial Scene Implementation
In October 22, 2026, Lingdong General, an embodied intelligence company incubated by iFlytek, released the LDB Embodied Intelligence General-Purpose Robot at the Developer Festival. The robot aims to address challenges caused by changes in product forms during warehouse sorting. By integrating the iFlyBot-WAM industrial world action model and the LindenBot-Harness execution organization system, the robot enhances flexibility and safety in industrial scenarios through fully self-researched technology. It has passed the national CR certification, providing verifiable safety and compliance basis for the industrial application of embodied intelligence.
PhoneBot: A Dual-Leg Robot Powered by a Repurposed Smartphone
PhoneBot, developed by UCLA's RoMeLa lab, is a low-cost, open-source dual-leg robot that uses a repurposed smartphone as its brain. It can walk, recover from falls, and perform basic interactions, all while using Wi-Fi for wireless control. The design leverages existing smartphone hardware to reduce costs and simplify development, making it accessible for research and education.
Deployments
Video｜Personal Robots Mass Production, 110 Humanoid Dancers Set World Record—Physical AI is Approaching Everyday Life
The article reports on the mass production of personal robots, with the 5000th unit being manufactured in under a month, achieving a production rate of one robot every 2.5 minutes. It also highlights the world record for 110 humanoid robots performing synchronized dance, showcasing the scalability and coordination of production-ready robots. Additionally, it discusses the deployment of robots in industrial settings, emphasizing real-world applications and the transition from lab research to practical use.
Zhiyuan Robotics Sets Guinness World Record: 100 Humanoid Robots Perform Synchronized Dance
Zhiyuan Robotics successfully challenged the Guinness World Records in Jinan, Shandong, with 110 Farview A3 humanoid robots performing synchronized dance, wave movements, and countdown displays. The demonstration showcased breakthroughs in large-scale coordination and spatial motion intelligence for mass-produced humanoid robots, utilizing self-developed LinkBots cluster control platform and UWB centimeter-level positioning technology to achieve millisecond-level synchronization.
Briefs
- IROS 2026 Best Paper in the Field | Visual SLAM Frequently Loses Track Underwater, This Team Uses Robotic Arm Suction Cups for LocalizationRoboSpeak — WeChat
- 110 Units of Farview A3 Full-Scale Humanoid Robots Synchronize to Set Guinness World RecordChinese robotics — Hardware and sensing
- PhoneBot: A Bipedal Robot Controlled by a Repurposed SmartphoneChinese robotics — Research news discovery
- Can a social reflex stop a robot but never restart it? Testing a dual-dwell safety gate (Webots)ROS Discourse — Community updates
- Zhiyuan Releases GenieStudio 2.0, Robots Have a New BattlefieldChinese robotics — Dataset and collection discovery
- Focus on 20 Core Humanoid Robot Chain Companies in the Next MonthChinese robotics — Hardware and sensing