THE COLLECTION
EVERY SIGNAL, SORTED.
Search the whole shelf by topic, tool, term, or plain old curiosity.
United States, Allen Institute for AI, Ai2, ToyBlocks, FoundationWorlds, Robotics, Simulation, Robot Learning, Manipulation, Isaac Lab, Reinforcement Learning, Motion Planning, Vision-Language-Action Models, Sim-to-Real Transfer, Open Source and Evaluation
ToyBlocks gave robot learning a small world. Keep the reality gap visible
Ai2 built an open tabletop world where robot policies can push, flip, rotate and stack blocks while researchers control the data and inspect failures. The small world is useful precisely because it does not pretend to be the whole one.
France, Hugging Face, LeRobot, Open Source Robotics, Interactive Rollouts, Remote Inference, Vision-Language-Action Models, Human Control, Emergency Stops, Network Failure, Action Logs and Physical Safety
LeRobot can take new instructions mid-motion. Put the stop boundary beside the prompt
LeRobot 0.7 lets an operator change a subtask while a policy remains loaded and lets a robot call a remote policy server. The prompt is becoming a control surface, so the independent stop circuit belongs next to it.
Japan, Fujitsu, Spatial World Models, Physical AI, Robotics, Service Robots, Security Robots, Scene Graphs, Human-Robot Collaboration, Robot Coordination, Safety Testing and Privacy
Fujitsu opened a spatial world model portal. Test the future before robots act on it
Fujitsu says its technology can combine fixed and robot cameras into a shared 3D scene graph, then imagine what people and machines may do next. Before that prediction becomes motion, operators need a test that treats uncertainty, occlusion and safe retreat as first-class evidence.
Embodied AI, Robotics, Xiaomi, Tsinghua University, University of Macau, Navigation, Manipulation, Vision-Language Models, Reinforced Self-Improvement, Recovery and Safety Testing
Xiaomi put navigation and manipulation in one robot brain. Keep the skill boundary visible
OmniBrain-0 gives a robot one high-level decision model for two very different jobs. The useful question is not whether the brain looks unified. It is whether every handoff, failure and recovery remains inspectable.
World Models, Compositional Generalization, Robotics, Planning, Benchmarks, Distribution Shift, Model Predictive Control, Open Source and Human Oversight
AnyGoal moved the finish line. World models did not carry the rules with them
AnyGoal keeps states, actions and physical rules familiar while changing how they combine. Across four controlled environments and eight goal-shifted variants, current world models often followed the behavior they had seen instead of the action they were given.
France, Open Robotics, Vision-Language-Action Models, Robot Memory, Manipulation, Public Training Data, Benchmark Evaluation and Human Safety
Kyutai put memory in the robot's action expert. The real room still gets a vote
REKI gives a 2.4B robot policy a cheap memory of its recent camera frames and makes repeated pose-switching less common in the team's tests. The architecture is clever. The next useful proof is a messy room, an independent operator and a recovery plan.
South Korea, Physical AI, Robotics, Manufacturing, Testbeds, Safety, Platforms and Industrial Policy
KAIST wants physical AI to be a platform, not a humanoid race. The test floor decides
KAIST is presenting physical AI as shared infrastructure for factories, healthcare, logistics and daily life. The useful measure will not be a robot sprint. It will be whether different machines can fail safely, share interfaces and survive repeatable tests in the real world.
Embodied AI, Robotics Data and Physical Safety
China's robot brain race runs on a thousand human demonstrations
Spirit AI forecasts a 2027 leap for general-purpose robots. Its more revealing evidence is a Beijing room full of people opening fridges, turning locks, and teaching machines how bodies move.
Robotics, Demonstration Learning and Human Control
A robot learned the new job from one video, then got a human check
Skild AI says S1 can attempt a new multistep task from a single visual demonstration without changing its weights. Fast instruction is useful. Reliable physical work still needs testing, recovery, and a stop button.
Autonomous Science and Physical Systems
Before the AI touches the robot arm, it has to prove the experiment is possible
Researchers built a computable representation of the laboratory so an agent can verify equipment, materials, dependencies, and safety constraints before physical work begins.
Robotics and Physical AI
A useful robot model needs an assembly line for learning
AWS and Nvidia are treating robot development as a continuous factory that generates experience, trains behavior, tests it in simulation, and starts the loop again.