On October 22, AgiBot officially open-sourced the fourth installment of the AGIBOT WORLD 2026 dataset, the "Physical Reasoning" dataset, which it describes as the first real-world Physical Agent dataset.
Geek+'s general-purpose humanoid robot Gino 1 will appear at CeMAT Asia 2026 from November 3 to 6, 2026, at Booth A3-2, Hall N1, Shanghai New International Expo Centre. The on-site demonstration will showcase multi-category picking and cross-task transitions, combining multimodal perception and force-control feedback to determine grasp positions and strategies in real time, and it is equipped with the self-developed Gravity embodied intelligence model; under full-body parallel control, peak throughput can reach 300UPH.
A CITIC Securities research report notes that institutions such as NVIDIA and Stanford are exploring the use of GPT-6 Astra for robot control, leveraging its visual understanding, spatial reasoning, and task planning capabilities to reduce reliance on specialized policy training. Astra has shown good generalization on embodied task datasets. CITIC Securities believes that in the near term, general-purpose models are more likely to divide and coordinate work with specialized models such as VLA and WAM, and that general-purpose models entering the embodied task field in 2027 will become the AI industry's next investment direction.
The study proposes The Trinity, a recursive self-improvement framework that couples world generation with Harness evolution through three roles: Creator, Explorer, and Evolver.
OSWAM proposes a world action model with one-step action generation. By constructing imagination tokens to initialize the diffusion process and providing physical priors and action-conditioned representations, it compresses the tens of denoising steps required by direct-policy world action models into a single step.
PhysEvolve proposes a verifier-grounded self-evolving framework that lets vision-language models learn causal physical reasoning from executable world programs they propose themselves and from simulation-generated tasks. In each round, a shared VLM proposes world programs and evidence schemas; a deterministic physics engine freezes each valid world and simulates matched baseline and body-elided counterfactual branches to generate discrete visual tasks; then exact verifier feedback updates the policy and curriculum via GRPO.
Rollback proposes a policy improvement paradigm that turns an action-conditioned world model from a one-way simulator into a backtrackable counterfactual engine: it returns to key decision points, branches to try different actions, and compares the outcomes, thereby distilling counterfactual results into action-level supervision without learning a separate value model.
Why it matters: The world model is used as a backtrackable counterfactual engine to provide fine-grained action-level supervision for policy improvement, and the improvement margins on three real-robot tasks are reported.
The study proposes OSMA-Bench, an obstacle-conditioned safe motion adaptation benchmark built on spatially matched scenario sets that differ only in obstacle type, and introduces the Semantic Sensitivity Index (SSI) to measure whether a robot's clearance and its distance-dependent speed and smoothness follow the caution ordering of obstacles.
Pollen Robotics showcased Microduck, a small bipedal robot positioned as a reinforcement learning development tool, at NVIDIA GTC Berlin. The robot was described as "a small bipedal platform for reinforcement learning," and the on-site team explained its development uses.
At an all-hands meeting made public on October 22, Figure AI announced that its fourth-generation humanoid robot, Figure 04, has completed design lock and entered the hardware verification and testing phase, and that a larger version of its in-house end-to-end vision-language-action model Helix 3.0 is expected to complete internal testing within a few weeks.
Teradyne Robotics and Sereact have formed a partnership under which Sereact's Cortex physical AI platform will drive Universal Robots robot arms for unstructured tasks such as returns handling and picking; the two parties have already delivered dual-arm systems made up of hundreds of UR robot arms to customer sites in Europe.
ETH Zurich's ANYmal quadruped robot solved a 10-pound giant Rubik's Cube using only its feet, three times, taking about 8 to 9 minutes each time. The demonstration combines computer vision, a traditional Rubik's Cube solving algorithm, and reinforcement learning to teach the robot to balance, apply force, and correct errors. This "pedipulation" (manipulation with feet) could enable future robots to open doors, turn valves, or move obstacles without hands.
ByteDance's Seed team updated its robotics hiring this month, extending roles from algorithms and simulation to data collection and robot hardware. Embodied data positions offer monthly pay of 55,000 to 85,000 yuan, and the data collection lead role starts above 70,000 yuan. The hiring covers Beijing and Shanghai and requires candidates to have experience in robot data collection, teleoperation, or on-site project management. Industry rumors say Seed has been collecting Ego data in China for several months.
HD Hyundai Robotics showcased its self-developed intelligent dual-arm robot at FIX 2026 and FABTECH 2026. It replicates the human upper body, with a five-finger dexterous hand at the end of each arm, a maximum payload of 7 kg per arm and a working radius of 630 mm. It runs on the company's self-developed Robot Foundation Model (RFM) and can autonomously perceive its environment and generate motion plans.
Enact Intelligence, a physical AI company headquartered in Zurich, Switzerland, has closed a €26.68 million (about $30 million) pre-seed funding round, with investors including redalpine, Notion Capital, and firstminute Capital.
Deep Blue AI published an article outlining the division of labor among VLA, WAM, and Harness in embodied systems: VLA turns observations and goals into actions, WAM uses future-state prediction to aid action learning, and Harness organizes subtasks, capability invocation, and execution feedback.
At the 140th Canton Fair, AgiBot Elf G2, Zhidong Future's "Xiaozhi," and Dongyi Technology's PHYBOT C-series bipedal humanoid robots moved from exhibition booths to actual roles, taking on tasks such as subway guidance, showroom presentations, and autonomous badminton and table tennis rallies, with service continuing until November 4. Dongyi Technology supports cross-scenario capability transfer with the PhyAgents decision-making agent in its Embodied PhyStack architecture and has launched a modular deployment kit, PhyAgents Hub. The industry is also exploring two commercialization paths: systematic foundations and positioning in vertical scenarios such as those of CloudMinds.
DEEP Robotics has released five research results built around its self-developed embodied brain, DEEPRobotics-Brain: the world model DR-WM-1.0, the embodied foundation model DR-VLA-2.0, the embodied agent DR-Agent, the policy optimization framework DR-APO, and the data closed-loop platform DR-Benchmark.
Ding Ning's team at Tsinghua University, together with researchers from Beijing University of Posts and Telecommunications, Peking University, and Nature Will, built HandsOn, a simulation platform and evaluation benchmark for cross-embodiment dexterous manipulation, containing 66 manually verified tasks and 18,000 human simulation teleoperation demonstrations.
Why it matters: Readers can learn why dexterous manipulation evaluation should distinguish execution from transfer, and which specific failure points current multi-task policies have on bimanual collaboration tasks.
DEEP Robotics has released five research results built around its self-developed embodied brain DEEPRobotics-Brain: the world model DR-WM-1.0, the embodied foundation model DR-VLA-2.0, the embodied agent DR-Agent, the policy optimization framework DR-APO, and the data closed-loop platform DR-Benchmark. Partners include seven universities, among them MIT, Tsinghua University, and Zhejiang University.
The MonoRace system from TU Delft's MAVLab in the Netherlands, relying only on a single front-facing camera and an IMU, defeated three world-champion-level pilots in a row at the 2025 Abu Dhabi Autonomous Drone Racing competition; the paper was published in Science Robotics on October 14, 2026.
Why it matters: The Delft team broke down how a monocular camera plus IMU divides perception, state estimation, and reinforcement-learning control in high-speed racing, which can be transferred to real onboard systems with limited compute.
A research team from Harvard, NVIDIA, Berkeley, MIT and Stanford proposes RAX (Robot Agent eXploration), which connects a general coding agent to robot perception and control interfaces, uses online exploration to write successful decisions into executable programs and accumulate them as reusable skills, while keeping model parameters frozen.
Why it matters: RAX uses a frozen model plus online exploration to generate reusable programs, leading by an average of 37.4 percentage points on five simulation benchmarks with zero action demonstrations, offering a path for robot learning different from scaling data and models.
The prospectus of Shenzhen Sengge Data Technology Co., Ltd. has been made public by the U.S. Securities and Exchange Commission (SEC), advancing its listing process. Its Sen-3D platform consists of Sen-Modeling high-precision 3D reconstruction and Sen-Creation physically consistent world model generation, and works with a structured 4D world state acquisition system to build a Real2Sim2Real closed-loop training pipeline. Official materials claim a Sim2Real zero-shot transfer success rate of over 80%; this approach still requires validation on more real-world tasks.
ProAS is a plug-and-play progress-aware advantage shaping framework that uses a dynamically evolving Agentic Reference Graph (ARG) to achieve fine-grained credit assignment, mitigating the problem in critic-free algorithms such as GRPO where trajectory-level advantages are broadcast equally to all actions.
The research team proposes the RynnVLA-Latent framework, which uses RynnLAM to learn continuous, 3D-grounded latent actions from visual changes without requiring native robot control signals. The method processes 30.1K hours of human, robot, and simulation video, generating 745.52M latent action descriptors, then uses flow matching to predict latent action chunks and adapt them to specific embodiments.
Why it matters: This work converts heterogeneous interaction videos into 3D latent action supervision, offering a reusable data pathway for VLA pretraining that bypasses the scarcity of action annotations.
HuGe is a family of humanoid generalist vision-language-action (VLA) models pretrained on over 2,000 hours of egocentric human and robot data. It predicts latent motion tokens to drive a pretrained whole-body controller and outputs hand-control signals.
A Microsoft team proposed the Code Video Model, in which a Coding Agent turns a prompt into a Three.js program, unifying geometric layout, object motion, camera trajectories, and event timing into an executable scene representation, then uses a program-previewed video to guide video generation.
Chelsea Finn shared Alexander Swerdlow's paper, which proposes generative value classifiers, a way of parameterizing Q-functions that does not require action conditioning. The authors believe this type of method can be combined with large-scale pretrained policies and can learn and generalize better from richer output signals. The paper link is alexanderswerdlow.github.io/gvc/.
MIT researchers have developed a paper-thin micro swimming robot driven by living muscle cells; the cells respond to flashes of light, enabling the robot to swim and steer. The research is seen as potentially advancing future biohybrid robots and micro-machines.
DexHoldem, a project involving the University of Hong Kong and Yisheng Technology, brings Texas Hold'em to real hardware, using a UR10e robotic arm and a Shadow dexterous hand to simultaneously evaluate dexterous manipulation, table-state perception, and closed-loop operation. It has been accepted to the NeurIPS 2026 Workshop.
Lingxin Qiaoshou and Xbotics have opened registration for the third dexterous hand online training camp, which starts on November 1 and runs for 8 weeks with 6–8 hours of commitment per week. The hands-on project is to recreate continuous in-hand ball rotation with the O20 in MuJoCo simulation, with O30 ball rotation as an extended example.
Registration is open for the RoboWorld Challenge 2026, which features three tracks: WorldNav, HA-VLN 2.0, and SafeDrive-VLA. They respectively test language-guided open-loop route planning, social navigation in dynamic crowds, and safe driving under conflicting instructions. The final submission deadline is November 30, 2026.
Authors from the Hong Kong University of Science and Technology propose World-to-Wrist VLA (W2-VLA), which uses global task context to predict future wrist latent variables and then guide fine action generation. The method achieves 98.5% average success rate on LIBERO, 60.71% and 18.21% on the Easy/Hard settings of RoboTwin 2.0, respectively, and outperforms baselines under standard and OOD conditions on three real-world tasks: tabletop cleanup, occluded placement, and dual-arm plug insertion.
Addressing the problem that world-model predictions may be built on outdated perceptions of the environment, SLAM-MER at CVPR 2026 uses recent keyframe temporal information and a retrievable 3D representation for monocular SLAM, while GaME in the same year continuously updates its 3D Gaussian representation and removes outdated keyframe observations.
Researchers from Shanghai AI Lab, Zhejiang University, and Shanghai Jiao Tong University propose EgoExplorer, a wearable capture system that uses a head-mounted GoPro HERO13, 5 VIVE Ultimate Trackers, and MANUS Metagloves or Wuji gloves to record 100 hours of human activity data, aligning first-person video, full-body motion reconstruction, finger movements, and tactile observations.
NVIDIA released an illustrated tutorial that uses a fictional startup building a warehouse humanoid robot as an example, connecting tools including Isaac Sim, Isaac Lab, GR00T, Cosmos, and Omniverse, and covering virtual environment setup, demonstration teaching, policy training and testing, handling the Sim2Real gap, and deploying AI to real hardware. The tutorial also explains the steps between an AI model deciding to move an arm and the motor actually executing it, and the author hopes to exchange with developers who have actually used these tools about how real workflows differ from the examples.
Sereact this week reached a partnership with Teradyne Robotics (including Universal Robots and Mobile Industrial Robots) to have the Cortex physical AI platform drive UR robotic arms in handling unstructured tasks such as returns sorting and picking; the two sides say they have already delivered hundreds of UR dual-arm robotic systems at customer sites in Europe.
Eka Robotics announced this week that it has closed a $150M Series A round led by Andreessen Horowitz (a16z) and F-Prime, with the funds to be used to scale its technology and customer deployments.
The Kuaishou Kling AI team, together with Nanyang Technological University and HKUST (Guangzhou), proposed MetaView, a diffusion-based single-image novel view synthesis framework, and has open-sourced the code, model weights, and an online demo. The method is built on the MM-DiT architecture, injects implicit geometric priors extracted by a feed-forward geometry-aware network through parallel attention streams, and uses metric depth to anchor scale-aware RoPE, mitigating geometric collapse and scale drift under large viewpoint changes.
DEEP Robotics has released five research results built around its self-developed embodied brain DEEPRobotics-Brain: the world model DR-WM-1.0, the embodied foundation model DR-VLA-2.0, the embodied agent DR-Agent, the policy optimization framework DR-APO, and the data closed-loop platform DR-Benchmark. The company says the work was carried out jointly with seven universities, including MIT, Tsinghua University, and Zhejiang University.