Researchers at Caltech and Stanford have unveiled HomeBody, a system that gives GPT Astra persistent spatial memory and the ability to execute compound humanoid actions for long-horizon tasks. In a kitchen demo, a Unitree G1 robot guided by Astra cleaned the room and retrieved remembered items from ambiguous requests. The team said the system requires no environment-specific training or additional policy learning. HomeBody replaces the trained intermediate VLA layer in conventional three-part architectures—System2 VLM for vision and instructions, System1 VLA for control commands, and System0 for coordinated motion. As frontier models such as Astra improve, the researchers argue trained VLAs are no longer necessary and System2 can directly manage reusable motion skills. A plug-and-play VLM calls a composable skill library....
Figure, the U.S. humanoid robot developer, has released Helix 2.5, its most advanced neural network model to date, and said it enabled zero-shot whole-body autonomy in 30 real homes. Figure called it the first demonstration at that scale of zero-shot whole-body behaviour generalisation for a humanoid robot. The company said a single pre-trained foundation model was applied to three types of behaviour: locomotion; manipulation of rigid and deformable objects; and bimanual coordination, as well as active perception. Figure said that, with task-specific data, architecture, training method and evaluation unchanged, pre-training on Index alone raised zero-shot recognition success to 56% from 9%. Index is Figure’s robot dataset project built for global scale. The model also validated a Human-to-Humanoid Robot Transfer Scaling...
What happens if robot model training data is expanded from 300 hours to 30,000 hours? Can robotic systems, similar to large language models, continuously gain new capabilities as datasets grow? AGIBOT’s newly released native World Action Model (WAM), GE-Act 2.0, systematically answers that question for the first time. Unlike mainstream approaches that build upon existing video generation models, every parameter of GE-Act 2.0 — including core modules for visual representation, future prediction and action forecasting — is initialized from scratch and trained end-to-end on embodied manipulation data. After pre-training, the model undergoes zero-shot real robot tests on unfamiliar scenes and unseen objects without fine-tuning or demonstration for benchmark tasks. The evaluation covers 100 atomic tasks, 20 skill categories and two...
Chinese robotics company Unitree said it has fully open-sourced an upgraded version of its general-purpose humanoid foundation model, UnifoLM-WLA-1.0, setting new state-of-the-art results on multiple global open-source embodied model benchmarks. The model can perform both tabletop manipulation and whole-body mobile manipulation with a single model, enabling cross-task and cross-end-effector generalization, Unitree said. It achieves a “one-model-driven, whole-body coordination” effect, the company added. Unitree said UnifoLM-WLA-1.0 was trained on large-scale general-purpose multimodal perception and understanding datasets and incorporates interaction-centric world modeling, significantly strengthening spatial perception and scene understanding. It outperformed several open-source competitors in China and overseas on multiple embodied reasoning benchmarks, with some capabilities comparable to mainstream closed-source models, according to the company. In real-robot tests, a single model completed...
Astribot, a Chinese robotics startup, has unveiled SmoothRL, an online reinforcement learning framework designed to train policies during asynchronous execution – a growing reality for embodied AI deployments where inference and action run at different frequencies. The work, led by the company’s foundation model team, addresses a fundamental mismatch: when a robot receives new action chunks before previous ones finish, conventional RL that updates on entire chunks ends up crediting or blaming actions that were never actually executed. The framework partitions each action chunk into three regions – committed (already fixed), execution (actually performed), and discarded (overwritten by later inference) – and backpropagates gradients only through the execution region. This ensures that optimisation mirrors the robot’s real experience. SmoothRL also...
Fei-Fei Li’s startup World Labs has released Atlas, a next-generation world model billed as the first multimodal system capable of pixel-level camera-controllable image and video generation paired with native 3D scene reconstruction. Trained from scratch on text, image, video and 3D data, the so-called “omni model” unifies four distinct modalities within a single unified architecture. Its flagship innovation, termed “spatial context”, grounds each input image in a precise 3D coordinate space with explicit camera pose parameters. By treating geometric camera controls as native inputs, Atlas eliminates reliance on textual prompts to describe camera motion, letting users specify exact viewing angles and trajectories to generate up to 1440p-resolution videos with a maximum one-minute duration. World Labs frames the upgrade as letting...
A mysterious Chinese artificial intelligence team has released a series of uncut, one-take videos showing humanoid robots performing household tasks with fluid, human-like coordination, reigniting online speculation that embodied intelligence might be nearing a “ChatGPT moment.” The clips, which went viral last week, depict robots adapting to external physical disturbances while carrying a water glass, tidying a living room in a single long-horizon sequence, collaborating to smooth a bedsheet, and carrying multiple items at once – all without visible teleoperation, according to the developers. The underlying model, internally codenamed MVP – short for “Make Veritable People” – is designed to endow machines with physical reasoning and autonomous decision-making rather than probabilistic imitation, the team said in response to inquiries. In...
Briefs
Shenzhen embodied AI startup KNOWIN raises Angel+++ round led by JD-linked fund
Shenzhen-based consumer embodied AI company KNOWIN said it has completed an Angel+++ round worth hundreds of millions of yuan, led by a JD.com-linked fund, with Loyal Valley Capital, Nanshan Strategic Emerging Investment and Walden International participating.
Proceeds will focus on expanding training data to improve GLOW's generalization across tasks, objects and home environments; deepening robot hardware R&D and advancing engineering validation and mass-production preparation for KNOWIN-X1; and recruiting senior embodied AI talent.
KNOWIN has now completed five financing rounds: seed, Angel, Angel+, Angel++ and Angel+++. Its Angel++ round raised 500 million yuan, with Matrix Partners China and Redpoint China among investors.
Founded in 2025, KNOWIN targets consumer home scenarios and aims to build embodied AI systems with physical bodies that can form a “perception-generation-execution-evolution” loop and interact continuously in real environments. Its folding dual-arm robot, KNOWIN-X1, is about 1.4 metres tall unfolded, 40 cm folded, weighs about 50 kg and has self-developed arms with a 6 kg payload.
China tactile robotics startup AGIMind Robotics raises tens of mln yuan in angel round
Shenzhen-based tactile robotics startup AGIMind Robotics said it has raised tens of millions of yuan in an angel funding round, with investors including Lihe Capital, Hefei Industrial Investment and other institutional investors.
The proceeds will be mainly used for production line construction, mass-production capacity optimization and R&D of tactile intelligent systems, the company said.
Founded in 2025, AGIMind Robotics focuses on tactile sensing and fine manipulation skills for robots. It has launched the Lux visuotactile sensor series.
Founder and CEO Zhang Yazhan holds a bachelor's degree from the University of Science and Technology of China and a PhD in robotics from the Hong Kong University of Science and Technology.
Manus parent Butterfly Effect raises over $500 mln in record China agent-native funding
Butterfly Effect, the parent of AI agent Manus, said it has completed a new funding round of more than $500 million, the largest ever raised by a China-based agent-native startup.
The round was led by Boyu Capital and IDG Capital, with existing investors Tencent, HongShan and ZhenFund also participating.
The funding news follows Manus's launch of Manus 2.0 and Cue, a multi-agent group chat app, on Sept. 29. It was the company's first major update since it formally resumed independent operations on Sept. 1.
Manus 2.0 uses Cascade, its in-house agent framework. The framework allows user-created projects to stay lightweight from the outset and invoke relevant specialized capabilities only when tasks actually require them.
Ashish Kumar, Former Tesla AI Lead Launches Robotics Startup
Ashish Kumar, who previously led Tesla's Optimus humanoid robot AI initiative, announced the founding of a new startup focused on developing non-humanoid robots. The venture is reportedly seeking to raise approximately $100 million in funding to innovate within the robotics industry.
Nvidia in talks to invest another $1 bln in humanoid robot startup Figure
Nvidia is in talks to invest an additional $1 billion in humanoid robot company Figure, extending a strategy that has shifted the chipmaker from financial investor to core technology supplier to the startup. Nvidia previously joined Figure's Series C round and provides hardware support for training and deploying Figure's Helix AI model. Nvidia CEO Jensen Huang has called the partnership a "robot flywheel": models are trained on Nvidia's Vera Rubin chips, validated on its Isaac Sim simulation platform and ultimately deployed on Nvidia GPUs inside the robots. The closed loop is accelerating Nvidia's strategic push into physical AI.
Zhongda Leader plans up to 530 mln yuan A-share private placement for embodied intelligence, automation projects
Zhongda Leader Intelligent Transmission Co Ltd disclosed a preliminary plan for a 2026 targeted private placement of A-shares, seeking to raise up to 530 million yuan, with proceeds mainly earmarked for two production projects—embodied intelligence joint modules and distributed intelligent actuator units for industrial automation—and an R&D center upgrade.
The company said the private placement is aimed at capturing opportunities in emerging industries such as embodied intelligence and seizing the window for domestic substitution of core components.
ByteDance preparing AI model for real-time spatial video
China's ByteDance is preparing to launch an artificial intelligence (AI) model capable of generating spatial video in real time, according to people familiar with the matter, entering the world-model field with potential uses in robots and autonomous systems and competing with Meta and Google parent Alphabet.
The new model builds on ByteDance's existing cinematic video-generation model Seedance, the sources said. It is designed to let users create interactive virtual worlds for livestreaming, short dramas and games.
The vision-based AI approach is viewed as critical to advances in robotics, gaming and self-driving cars. The sources spoke on condition of anonymity because the plans are not public.
Amazon to invest over $100 mln in Indiana robotics manufacturing plant
Amazon.com Inc said it plans to invest more than $100 million to build an advanced robotics manufacturing plant in Greenwood, Indiana, with production expected to begin in 2028. The facility will make warehouse robots and supporting automation equipment for Amazon's North American fulfillment network.
The plant will integrate precision component machining, robotic welding, automated powder coating and final assembly, and will rely on AWS cloud and smart manufacturing systems, Amazon said. All completed robots and components produced there will be supplied to Amazon's North American fulfillment centers and robotics operations network.
RobotHOT brings China & Global Robotics Insights. Explore robotics, embodied‑intelligence updates, startup news and industry analysis for professionals worldwide.
the 2026 Bund Summit