The paper proposes KITE (Kinematic Interaction Transfer across Embodiments), which splits manipulation into embodiment-agnostic interaction intent prediction and embodiment-specific kinematic decoding, connected through an interaction intent representation based on contact patterns.
ZeroDex is a zero-shot long-horizon dexterous manipulation framework that grounds language instructions into executable 3D task plans from calibrated multi-view RGB images. It leverages the 2D grounding ability VLMs already possess and lifts it to the metric 3D quantities manipulation requires (contact points, placement targets, tool-motion anchors) through multi-view fusion; execution is driven by these 3D groundings rather than a learned policy, requiring no demonstrations or post-training for a specific task or embodiment.
Researchers propose an interpretability pipeline to analyze what a fine-tuned vision-language-action (VLA) policy is computing internally. On pi0.5, injecting task A's activations into the action expert while the policy views task B's scene caused 99.6% of episodes (n=1968) to execute task A's motor trajectory, indicating the injected program is bound to absolute workspace coordinates and ignores the visible scene.
Why it matters: The study uses activation injection to break down the internal computation of VLA policies, offering a mechanistic explanation for understanding their perturbation brittleness.
AdaDex is a goal-conditioned reinforcement learning framework that turns a few seconds of human hand-object trajectory (from motion capture data or HOI reconstruction from in-the-wild RGB(D) video) into generalizable dexterous skills. Its core, Hybrid Dexterous Mimicking, trains a shared policy across two sets of massively parallel environments: environments with dense motion tracking objectives inject the human prior, while environments with noised initial states and goals and a sparse object-success reward force the skill to work even when it deviates from the reference trajectory.
PhysCanvas is a state-consistent visuo-tactile world model for contact-rich robot manipulation that unifies multi-view RGB, tactile images, marker displacement, task state, and actions within a single vision space, sharing a video VAE and a pretrained generative backbone.
The research proposes GPT-as-Tactile-Policy, which summarizes expert demonstrations into task guidance and tactile references, enabling GPT-6-Astra to combine multimodal observations and interaction history to generate control commands and adjust actions based on contact feedback.
The paper proposes TaskWAM, a closed-loop, stage-aware framework that uses Mode Tokens to represent fine-grained task requirements and State Tokens to track execution progress, forming a structured task state that steers world-action model generation toward futures matching the current task, addressing the Stage Confusion phenomenon in long-horizon manipulation.
The paper systematically evaluates GPT-6 Astra's closed-loop robot manipulation capabilities using a lightweight agent execution framework, finding a 26.92% success rate on RoboDojo and 76.0% on RoboTwin 2.0, but a clear gap between semantic reasoning and physical execution.
The paper proposes a two-stage framework that learns complementary predictive representations from multi-view video clips to enhance embodied manipulation policies. In the pretraining stage, a frozen 3D foundation model provides dense geometric features, which are compressed into predictive geometric representations and optimized to preserve scene structure and its temporal evolution; action supervision is then applied to geometric transitions to extract a motor representation that emphasizes action-relevant transitions. In the downstream post-training stage, predictive geometry is transferred through geometry queries, and the motor representation guides intermediate features along the action-generation pathway. Future observations serve only as privileged supervision during training and are not needed at inference time.
An arXiv paper presents LocoPilot, a single-operator bilateral teleoperation and data-collection framework deployed on a 12-DoF Unitree Go2 quadruped equipped with a 6-DoF AgileX Piper L manipulator, an Intel RealSense D435 wrist camera, and a D405 head camera.
The paper proposes the HuTact framework, which uses large-scale human hand-object interaction (HOI) data to pretrain tactile-aware hand representations. The pretraining stage requires no real tactile signals, and the model is then adapted with a small amount of unlabeled robot interaction data.
EventSync proposes a hierarchical framework with event-driven memory updates to address history-dependent manipulation tasks where the current observation alone is insufficient to determine the next action. A vision-language model generates subgoals based on the task plan and interaction history, while a vision-language-action model handles execution. A lightweight Event Monitor combines visual observations with end-effector trajectories to determine whether the current subgoal is complete; only after obtaining completion evidence does it write the completed subgoal into the history and trigger subsequent reasoning.
A CoRL 2026 paper proposes Advantage-Weighted Latent Steering (AWLS), an offline post-training method that improves frozen one-step generative manipulation policies without updating their decoders.
The paper proposes replacing the wrist camera, pose tracker, and their separate data streams in the UMI glove setup with a single monocular camera (USD 122) worn on the collector's head. It reads printed fiducial markers on an electronics-free glove (USD 22) and paper tiles in the workspace, producing the 6-DoF gripper pose and jaw width from the same frames the policy will see. At deployment, the same camera is mounted on a low-cost (USD 120) robotic arm beside the robot and mimics the collector's head motion based on the robot's joint states.
The paper introduces JARVIS, a method-agnostic benchmark for autonomous execution guided by visual manuals without additional language instructions, comprising 236,679 LEGO assembly tasks across 8,987 scenes, plus a curated JARVIS-v1 evaluation suite of 50 tasks.
The Tactile-ETA system brings tactile experience into embodied task agents through in-context learning, completing fine manipulation without updating model parameters. Across four simulated manipulation tasks, adding tactile observations to the same two expert demonstrations raised average success from 15.75% to 24.00%; on a real robot task, vision- and tactile-aligned demonstrations raised success from 0% to 60%. Experiments also show that performance depends on the sensory content and representation of the demonstrations, and different agent models differ markedly on the Lift Bottle task.
HATCH is an open-specification, physically replicable, sensor-scored, and unrehearsable robot manipulation benchmark, accepted to the 5th Open Robot Software Workshop at IROS 2026 (2 pages, 1 table). Tasks span five skill categories and three difficulty levels, and a sensor that confirms completion of one task unlocks the next, chaining an entire room into a single long-horizon evaluation.
The paper proposes an autonomous bricklaying concept system based on a Boston Dynamics Spot quadruped robot with a manipulator arm, operating between a supply area and an assembly area, using AprilTag localization and autowalk mission navigation, and exporting target brick poses directly from CAD software for the robot to execute.
EmbodyHOI runs a differentiable inverse kinematics (IK) solver for a specific humanoid embodiment inside the sampling loop of a frozen flow-matching hand-object interaction (HOI) trajectory generator, placing embodiment reachability constraints within the generation process rather than in post-hoc retargeting.
A study proposes a plug-and-play auxiliary loss called score-decorrelation to address the problem that diffusion policies with linearly composed scores tend to have convergent component predictions that are hard to interpret in multitask robot learning. The loss encourages specialization by penalizing overlap between component noise predictions, with almost no architectural changes and very low computational overhead. Experiments on multiple benchmarks show that the method consistently reduces duplicate counting between components, improves task decomposition, and produces behaviorally meaningful specialization without compromising multitask performance.
DexGraspFlow proposes a conditional link-level SE(3) flow-matching framework for cross-embodiment dexterous grasp generation. It combines geometric relations with kinematic topology, uses a dual-stream DiT to model link-link coordination and link-object interaction, and provides explicit structural information through attention biases derived from kinematic-tree distances.
The article analyzes why the dexterous hand industry is collectively reducing degrees of freedom: manufacturers are now talking about lifespan, cycle time, and unit price rather than making hands more human-like. A new hand from a leading US robotics company has only four fingers and 13 degrees of freedom, while two new domestic products have 22 and 20 degrees of freedom, compared with benchmark products from just over a year ago at 38, 37, and 42.
AgiBot has partnered with Wuhan Tourism Group and Zhongyun Weixun to deploy the Expedition A3 Ultra full-size bipedal humanoid robot for real-world commercial operations at the "Jiangyu" bridge-pier film camp on the Hankou riverbank. As of October 19, the Expedition A3 Ultra had run stably for 18 days, completed 827 orders cumulatively, and achieved an overall grasp success rate of over 97%; this weekend, single-day orders exceeded 100 with a grasp success rate of 100%.
A Grip Sensor board gives basic PWM servos a sense of touch by measuring servo current to estimate grip force. A status LED indicates whether an object has been grasped, a Teach button lets users teach the grip strength, and an open-drain status output can communicate with existing MCUs. In the video demo, the servo grips a box; when the box drops, the board detects it and notifies the MCU, while also preventing the servo from gripping too tightly and overheating. The author has made 5 boards for testing, available by direct message.
Nature Will recently released a demonstration video of the Natura-1 model, in which robotic hands, without human assistance, completed tasks such as wrapping wontons, packing express deliveries, brewing tea, applying phone screen protectors, and writing Chinese calligraphy with a brush. Natura-1 is a 50-billion-parameter mixture-of-experts (MoE) model, with data processing covering nearly 1 million hours of first-person human interaction data. It can drive a 20-degree-of-freedom dexterous hand within a 100-millisecond control loop and follow complex natural language instructions to generalize in unseen scenarios.
Dream Labs released Dream-1 yesterday. A Reddit user who deployed it on a robotic arm said it is fast and can perform tasks zero-shot, but it also exhibits random behavior, such as repeatedly grabbing a red block when asked to move an apple. The model has been made public for people to try.
Dream Labs released Dream-1 yesterday. A user deployed it on a robotic arm, saying it is very fast and can perform tasks zero-shot. In testing, when asked to move an apple, it instead grabbed a red block and did not stop, showing occasional random behavior. The publisher has made Dream-1 publicly available for people to try.
Warehouse robotics company Exotec announced the completion of two Skypod system deployments in Japan, with partners food distributor Mitsubishi Shokuhin and industrial goods distributor Trusco Nakayama; the projects were implemented in partnership with IHI Logistics Industrial Systems.
Eka Robotics has completed a $150 million Series A funding round. Pieter Abbeel reposted the news and congratulated the team. The quoted post states that the company pursues both manipulation performance and generality with its Vision-Force-Action (VFA) model, arguing that force is the language of manipulation, reinforcement learning is the path to scaling dexterous manipulation, and fast, reliable physical execution is the bottleneck to be solved; it also plans to use the funding to scale up its models and accelerate customer deployment.
The Xbotics Embodied AI Lab has released Lecture 6 of its 21-lecture series, explaining the complete perception pipeline from RGB-D observations to grasp poses, with accompanying code in the GitHub repository Xbotics-Embodied-AI-club/Xbotics-Embodied-AI-Handbook.
Plus One Robotics demonstrated AI vision depalletizing and palletizing cells live at PACK EXPO 2026 in Chicago, combining Torfresma's TRF Compact system with ABB robots in a modular workstation.
The PANVIS STAR® neurointerventional robot completed full-workflow validation of five major procedure categories in a porcine model: cerebral angiography and treatment access establishment, aneurysm coil embolization, flow diverter placement, vertebral artery stenting, and carotid artery stent implantation. All procedures were successful, with no device-related complications or vascular injury. The study was published in the Journal of Robotic Surgery.
A 6-DOF robot hand adapted from the open-source HRI Hand design, set up for servo drive, using a rigid PLA structure and flexible TPU fingertips to improve contact and grasping, aimed at robotics projects, education, and grasping experiments. Editable STEP files and STL files for 3D printing are provided and available on the Cults platform; only digital files are sold, not including servos, electronics, or assembly hardware.
A team from Nanyang Technological University used a system combining the AgileX PiPER-X robotic arm and the Zhishen Technology Tongchui M1 quadruped robot to win the championship and the Best Autonomous Capability Award at the IROS 2026 LRC legged robot challenge, with the task scenario being local hazardous chemical label recognition.
RoboValue introduces a fine-grained sim-and-real benchmark for unified evaluation of robotic value models, assessing four complementary capability dimensions through shared interfaces: task-state understanding, temporal progress monitoring, failure and recovery reasoning, and value consistency.
The article analyzes why robots are entering logistics sorting scenarios in 2026: sorting is currently the only scenario that simultaneously has orders, repeat purchases, data, and tolerance for failure. Real data generated in the feeding process can flow back, and the gap between a 3% return rate and the 2% rate for manual labor is only a cost issue, not an accident.
This paper systematically studies the mechanisms, scaling behavior, and generalization of physical recursive self-improvement (RSI) in embodied agents, proposing PhysRSI-eval, which quantifies physical RSI across five dimensions: first-attempt performance, efficiency, success rate, budget, and improvement from failure, and evaluates multiple VLMs on 18 manipulation tasks under a fixed interaction budget.
The research team proposes HandsOn, a composable cross-embodiment dexterous manipulation simulation platform and benchmark supporting 5 robotic arms, 6 dexterous hands, and about 200 rigid, articulated, and deformable assets. The benchmark contains 18,000 human demonstrations and 66 human-verified tasks, of which Core-50 is used for demonstration-based learning and execution tests, while Transfer-16 evaluates asset and relation transfer, skill composition, and deformable manipulation without target-task demonstrations.
The paper proposes the collective experience learning (CEL) paradigm, in which deployed robots share execution experience to improve a common policy without collecting additional expert demonstrations or corrective actions. It uses failure-guided experience retrieval fine-tuning (FERT) to retrieve relevant trajectory segments from the fleet's shared successful experience as supervision, and a fault-tolerant asynchronous training system (FATS) to coordinate experience collection, policy optimization, and distribution.
VLDA proposes modeling task-relevant state transitions with compact Delta variables, replacing VLA's direct observation-to-action mapping and the full future-state prediction of world-action models. It learns Delta under a unified objective via LT-Filter and GT-Anchor, and DDAR injects Delta into the action denoising process as a structured seed. Experiments implemented in the open-source training framework LoongForge show that VLDA improves long-horizon consistency, robustness, and generalization.