AgiBot G2

Headlines · 3

1. Oct 24
   China’s Embodied AI Robot Exporters Adopt Lighter Delivery Models

   TMTPost reports that China’s robot exports reached 24.85 billion yuan in the first half of 2026, up 359% year over year, according to customs data. However, industry research cited by TMTPost says entertainment, education and related purchases account for more than 60% of exported robots, while only 13% enter smart manufacturing and 5% serve warehousing and logistics. Chinese embodied-AI companies are responding with cloud-based teleoperation, modular hardware, shared maintenance networks and robot-as-a-service models to reduce overseas deployment and support costs.

   钛媒体 · 🔥 0
2. Oct 19
   SenseTime study: GPT-6 Astra aces simulated embodied tests, stumbles on real robots

   A SenseTime SenseNova team, with Daxiao Robotics, Ropedia, NTU and the University of Tokyo, published "Has GPT-6-Astra Achieved Embodied Intelligence?", a systematic evaluation of OpenAI's GPT-6 Astra across 12 simulated benchmarks and 8 real robot platforms. With in-context learning and up to a 10x action-step budget, the model scored 100% on LIBERO-Long++, 99.3% on LIBERO-Pro, 98.8% on SIMPLE and lifted RoboDojo from 27.6% to 81.9%, but reached only 37.0% on the long-horizon BEHAVIOR benchmark and 28.9% on LHPR-VLN navigation. On real hardware it succeeded 17 of 18 times at rigid pick-and-place, yet failed all 11 plug-insertion attempts across three platforms; placing a pen in a holder took over four hours and cost $55, and in a cooking task it decided an egg was done but turned off the stove about 90 seconds later, burning it. The report concludes GPT-6 Astra has largely crossed the embodied-intelligence threshold but remains far from practical deployment, and presents SenseNova-RoboRSI, a harness that improves from execution feedback without updating model weights.

   36氪 人工智能 · 🔥 0
3. Oct 18
   Robot fleets learn from shared experience, reaching 98.3% on phone packaging

   An arXiv preprint from a team at Tsinghua University, AgiBot and Infinigence AI introduces collective experience learning (CEL), a paradigm in which deployed robots share execution experience to improve one common policy. Its algorithm, failure-guided experience retrieval fine-tuning (FERT), uses the context of a failed execution to retrieve relevant successful trajectory segments from the fleet's shared experience and trains on their actions, requiring no extra expert demonstrations or corrective actions — humans only label success or failure and reset the environment. A fault-tolerant asynchronous training system (FATS) keeps collection, optimization and distribution running through individual robot interruptions. The authors report 93.3–99.2% success across five real-world manipulation tasks over 120 trials each on robots excluded from training data collection, and a rise from 79.2% to 98.3% on multi-stage phone packaging when learning with 96 robots, while noting evaluation was limited to five task-specific policies on a single platform.

   Hugging Face · Papers · 🔥 0

Experience and discussion from the community

Share my AgiBot G2 experienceAsk about AgiBot G2