Infinigence AI, together with Tsinghua and AgiBot, proposes collective experience learning CEL, with 98.3% success rate validated on 96 robots
This week embodied intelligence models were released densely: Google DeepMind launched the real-time interactive general world model Genie 4, and AgiBot, Qiyuan, Tuoyuan Zhihui and others also successively updated embodied models and simulators. At the industry level, Zhongke Huiling completed a round A++ financing of several hundred million yuan, and Dream Labs obtained a 90 million USD seed round, with the funds directed toward mass-production delivery and productization. On the research side, Infinigence AI, together with Tsinghua and AgiBot, proposes collective experience learning CEL, raising the task success rate to 98.3% on 96 robots.
This week embodied intelligence models were released densely: Google DeepMind launched the real-time interactive general world model Genie 4, and AgiBot, Qiyuan, Tuoyuan Zhihui and others also successively updated embodied models and simulators. At the industry level, Zhongke Huiling completed a round A++ financing of several hundred million yuan, and Dream Labs obtained a 90 million USD seed round, with the funds directed toward mass-production delivery and productization. On the research side, Infinigence AI, together with Tsinghua and AgiBot, proposes collective experience learning CEL, raising the task success rate to 98.3% on 96 robots.
Highlights
Google DeepMind releases world model Genie 4, supporting real-time interaction and native audio
Google DeepMind released its new-generation general world model Genie 4 on the evening of the 28th. It can generate and modify interactive scenes in real time based on prompts, sustaining a single coherent world for about 2 minutes at roughly 24 frames per second.
36氪 · 微信公众号Source ↗Harvard, NVIDIA, and others propose RAX: robot agent online exploration, leading by an average of 37.4 points across five benchmarks
A research team from Harvard, NVIDIA, Berkeley, MIT, and Stanford proposed RAX (Robot Agent eXploration), which connects a general coding agent to robot perception and control interfaces, writes successful decisions into executable programs through online exploration and accumulates them as skills, while model parameters remain frozen.
CVer · 微信公众号Source ↗AgiBot releases embodied model GO-3 Preview, connecting full-pipeline scalable learning from pretraining to post-deployment
AgiBot released the embodied model GO-3 Preview, focused on full-pipeline scalable learning across pretraining, post-training, and post-deployment, and ranked first in three benchmarks: RoboDojo, RoboTwin, and RoboColiseum.
机器之心 · 微信公众号Source ↗
Model releases/updates
World models and embodied models became the main thread of this week's updates, with multiple vendors advancing simultaneously on interactive generation, full-pipeline learning, and open-source action models.
Google DeepMind releases world model Genie 4, supporting real-time interaction and native audio
Google DeepMind released its new-generation general world model Genie 4 on the evening of the 28th. It can generate and modify interactive scenes in real time based on prompts, sustaining a single coherent world for about 2 minutes at roughly 24 frames per second.
36氪 · 微信公众号Source ↗AgiBot releases embodied model GO-3 Preview, connecting full-pipeline scalable learning from pretraining to post-deployment
AgiBot released the embodied model GO-3 Preview, focused on full-pipeline scalable learning across pretraining, post-training, and post-deployment, and ranked first in three benchmarks: RoboDojo, RoboTwin, and RoboColiseum.
机器之心 · 微信公众号Source ↗Qiyuan releases PrimeSim world model simulator, topping EWMBench and WorldSimProbe
The Qiyuan Embodied Research Center released PrimeSim, a world model simulator for robotic manipulation tasks, built by adapting a pretrained video generation model through two stages of mid-training and post-training. It ranked first on EWMBench with a composite score of 6.09, and took first place in three of five WorldSimProbe evaluations: global trajectory coverage, action-source behavior preservation, and interaction dynamics.
机器之心 · 微信公众号Source ↗X-Era AI (Tuoyuan Zhihui) open-sources 4D world action model E0.5, with real-robot success rate surpassing π0.5
X-Era AI (Tuoyuan Zhihui) open-sourced the native 4D world action model E0.5, achieving overall average success rates of 85.0% and 88.3% on the AgileX Piper dual-arm robot and the Zerith H1 humanoid robot, respectively, higher than π0.5 by 13.3 and 26.7 percentage points.
机器之心 · 微信公众号Source ↗PokeBot releases embodied API RoboAPI, world-action model Poke-Aurora debuts
PokeBot released its first model product RoboAPI in the early hours of October 30, a publicly accessible, out-of-the-box embodied API. After developers connect and adapt a robotic arm, they need neither to collect data nor train models; a single prompt command can execute tasks that were not specifically trained for.
机器之心 · 微信公众号Source ↗AgiBot releases WITA-OMNI interaction model and Lingqu OS 1.0, to be fully open-sourced by year-end
At its 2026 Innovation Day, AgiBot released the embodied omni-modal end-to-end interaction large model WITA-OMNI, which ranked first in four benchmarks—Daily-Omni, TimeLens-Bench, EmbSpatialBench, and MMAU—with an average response time of about 600 milliseconds.
机器人前瞻 · 微信公众号Source ↗
Industry news
Financing and productization are accelerating, and discussion continues on general large models taking over robot operation.
Astra shockwave and the embodied industry's 50 days
In the 50 days since the release of GPT-6 Astra, the embodied industry has conducted intensive evaluations and discussions around it, and general large models have begun to directly take over robot operation. Tests by SenseTime together with multiple institutions concluded that Astra is basically capable of completing embodied tasks, but still has an obvious gap from practical use in real environments, underperforming in millimeter-level contact, dynamic control, and response speed. The article also reviews industry uses such as using Astra directly for policies, reviewing VLA actions, building simulation environments, and data annotation, as well as chain reactions including cooling financing and major companies stepping up data collection.
钛媒体Source ↗Zhongke Huiling completes several hundred million yuan A++ round, advancing embodied intelligence mass-production delivery
Zhongke Huiling completed an A++ round of financing in the several hundred million yuan range, jointly invested by Zhilai Investment, Yonghua Investment, Maotian Investment, Yuanhe Chenkun, Guoke Kezhi, Luning Investment, and others. The funds will be used for product mass-production delivery, R&D iteration of embodied brain technology, and market expansion. The company's self-developed CASBOT PA perception-to-action model has been validated at actual workstations in a shoe factory in Putian, achieving a takt time of 7 seconds per unit for the outsole stamping process and 18 seconds per pair for sole pressing. In 2026, its total on-hand orders for embodied shoe-making robots exceeded 400 million yuan, with equipment scheduling reaching the scale of thousands of units.
中科慧灵 · 微信公众号Source ↗Dream Labs raises $90 million seed round and launches Dream API and Dream-1 model
Dream Labs announced the completion of a $90 million seed round led by B Capital, with participation from Kindred Ventures and others. The company also launched Dream API, which can directly convert robot camera images and natural language commands into motor commands, and said its flagship model Dream-1 is trained on videos of everyday human tasks and can zero-shot control any robot equipped with a gripper.
Robotics & Automation NewsSource ↗
Paper research
New results have emerged in directions including collective experience learning, cross-embodiment control and touch, and navigation foundation models, with evaluation benchmarks improved in parallel.
Infinigence AI, together with Tsinghua and AgiBot, proposes collective experience learning CEL, with 98.3% success rate validated on 96 robots
Infinigence AI, together with Tsinghua University and AgiBot, proposed the collective experience learning (CEL) paradigm and completed prototype validation on 96 robots. In a phone packaging task, the policy trained with CEL completed 120 tests on two same-model robots that did not participate in training, with the success rate rising from 79.2% to 98.3%; when the number of robots participating in learning increased from 16 to 96, the test success rate rose from 86.7% to 98.3%.
3D视觉工坊 · 微信公众号Source ↗Harvard, NVIDIA, and others propose RAX: robot agent online exploration, leading by an average of 37.4 points across five benchmarks
A research team from Harvard, NVIDIA, Berkeley, MIT, and Stanford proposed RAX (Robot Agent eXploration), which connects a general coding agent to robot perception and control interfaces, writes successful decisions into executable programs through online exploration and accumulates them as skills, while model parameters remain frozen.
CVer · 微信公众号Source ↗TU Delft's MonoRace achieves autonomous drone racing at 100 km/h using only a single front-facing camera
The MonoRace system from the MAVLab at Delft University of Technology in the Netherlands relies only on a single front-facing camera and an IMU, and consecutively defeated three world-champion-level pilots at the 2025 Abu Dhabi Autonomous Drone Racing competition. The paper was published in Science Robotics on October 14, 2026.
深蓝AI · 微信公众号Source ↗WorldOracle releases action-conditioned world model benchmark; seven models show deviations between prediction and real execution
Princeton University, UT Austin, the University of Toronto, NVIDIA, and AgiBot jointly launched WorldOracle, an action-conditioned world model benchmark that compares model predictions with real execution using the same initial frame and action sequence.
The Robot ReportSource ↗X-WBC: a cross-embodiment foundation model for humanoid whole-body control
X-WBC is a cross-embodiment humanoid whole-body control foundation framework that separates relatively shared human motion semantics from embodiment-specific physical execution. It uses human-centric command tokens to align full-body human motion, robot reference motion, and sparse VR observations; a causal Transformer learns reusable temporal structure from mixed multi-robot rollouts, and lightweight robot-specific modules then map the shared representation to each robot's proprioception and action space.
arXiv cs.ROSource ↗HKU open-sources GenSfM: jointly recovering 3D objects and camera poses from four uncalibrated images
A team at the University of Hong Kong open-sourced GenSfM, which uses a native 3D diffusion model to jointly denoise object shape and per-view camera poses within a single multimodal sequence, requiring only 2 to 4 uncalibrated images and no camera intrinsics or extrinsics. Evaluated on Toys4k and GSO, the paper improves novel-view PSNR by 3.2–4.5 dB over the multi-view baseline TRELLIS-M, reduces median Chamfer distance by about 64%–74%, and achieves pose Acc@5° above 96%.
3D视觉工坊 · 微信公众号Source ↗RynnWorld-Latent: a cross-embodiment world model conditioned on general latent actions
The research team proposed RynnWorld-Latent, a cross-embodiment video world model that uses RynnLAM's 3D-grounded latent actions as a shared conditioning interface, and can be jointly pretrained on human interaction videos, robot demonstrations, and simulation data.
HuggingFace Daily PapersSource ↗Render-free event-frame simulation accelerates reinforcement learning for autonomous robot navigation
The authors proposed an analytical event simulator that directly synthesizes event streams from scene geometry and ego-motion without image rendering, generating 64x64 event observations in just 17 microseconds on a single GPU, more than 400 times faster than advanced rendering pipelines. An event-camera reinforcement learning obstacle-avoidance policy trained with this simulator converges in under 30 minutes, three orders of magnitude faster than training with rendered events, and can zero-shot transfer to Isaac Sim ray-traced event simulation.
arXiv cs.ROSource ↗FTP-1: a general foundation tactile policy across tactile sensors
The research team proposes FTP-1, a general-purpose foundational tactile policy pretrained across multiple tactile sensors and proprioception, which uses heterogeneous encoders to map image-, array-, and state-type tactile signals into unified morphology-aware latent tokens, jointly modeled by shared tactile Transformer experts.
arXiv cs.ROSource ↗RepresentNav: a navigation foundation model built on a self-constructed representation space
The authors propose the RepresentNav navigation foundation model, which uses Representation Residuals, a three-layer transformer, to aggregate full-depth features from multiple visual foundation models and construct a representation space driven by the policy's own objectives.
arXiv cs.ROSource ↗RoboCoach: using a world model to locate weak subtasks, with targeted teaching to improve long-horizon task success rates
The Tsinghua University team of Zhang Changshui and Cui Sen proposes the embodied intelligence system RoboCoach, which uses the world model CoachWorld to run long-horizon tasks in a closed loop in imagination, locate the first timed-out weak subtask and request targeted human teaching for it, updating only the corresponding skill expert. Under a fixed demonstration budget, the Franka success rate rises from 13.3% to 75.0%, and AgileX from 40.0% to 83.8%, while the uniform-collection baseline reaches only 30.0% and 47.5% respectively. The average success rate on four unseen new task routes is 35.0%, versus 0% for the shared-policy baseline.
机器之心 · 微信公众号Source ↗