Skip to content
#Other
  1. 雷科技78

    WorldOracle world-model evaluation platform launched, using real-robot testing to fact-check the industry

    On October 16, WorldOracle, an action-driven world-model evaluation platform jointly completed by researchers from AgiBot, NVIDIA, Princeton University, the University of Texas at Austin, and the University of Toronto, was officially released, opening evaluation submissions to developers.

    Why it matters: The evaluation compares world-model predictions frame by frame against real execution records, revealing a common problem in which robotic arm motion is replicated but object states are systematically hallucinated, providing concrete reference points for judging model usability.

  1. euRobotics76

    FCC bans new imports of foreign advanced robots; euRobotics responds with open letter

    The US Federal Communications Commission (FCC) has banned new imports of foreign-made advanced robots, and the euRobotics Board has sent an open letter to the European Commission on the matter. The stated reasons for the ban include national security threats, concerns about cyberattacks, and considerations of reshoring the supply chain; euRobotics believes it will both raise the threshold for European exports to the US and may divert Chinese imports toward the European market.

    Why it matters: The article sets out the three reasons for the US import ban and its two-way impact on Europe, and also gives the specific demands of the European robotics community to the European Commission.

  1. Physical Intelligence78

    Physical Intelligence proposes MEM, enabling π0.6's VLA to remember long-horizon tasks of up to 15 minutes

    Physical Intelligence proposed a method called Multi-scale Embodied Memory (MEM) that provides vision-language-action models (VLAs) with both short-term and long-term memory, enabling the model to complete tasks lasting up to 15 minutes, such as cleaning an entire kitchen or making a grilled cheese sandwich from scratch.

    Why it matters: The article identifies memory mechanisms as the bottleneck for long-horizon tasks and presents a dual-scale design that stores video for the short term and language for the long term, an approach that can be transferred to other work.

  1. Physical Intelligence73

    Physical Intelligence: VLA models show human-to-robot transfer as robot data scale grows

    Physical Intelligence research found that in generalization scenarios where only human demonstrations appear, fine-tuning π0.5 with human first-person videos mixed into robot data alone improved policy performance by about 2x, without using any explicit transfer learning mechanism.

    Why it matters: Experiments show that merely mixing human videos into fine-tuning improves the policy by about 2x on scenarios with purely human demonstrations, and this capability emerges as pretraining robot data scale grows.

You’re all caught up