Embodied AI

TL;DR for operators A robot can predict a plausible future and still take the wrong action. The operational question is therefore not simply how accurately a system models what happens next, but how that prediction is represented and connected to control. The tutorial distinguishes world models, which predict future task-relevant observations or states under actions, from world action models, which couple future prediction with action generation. It then turns this distinction into an architecture map: predict raw observations or compact states; expose the future explicitly or keep it latent; connect prediction to action through a separate controller, predictive features, joint generation, or auxiliary training. ...

TL;DR for operators A familiar robotics design question is where to draw the boundary between understanding and control: should one learned system interpret the scene, choose where to go next, and explain that choice, or should those functions remain distributed across separate perception and planning modules? VISOR1 tests a middle position. Its 3B model combines language understanding, visual-spatial reasoning, target recognition, and high-level destination selection, while a conventional planner handles the low-level movement. It is not the strongest navigator on the leaderboard, but on OVON its post-trained model records a 21.70% success rate on Val Seen and 22.00% on Val Unseen, an unusually small seen-to-unseen change even though stronger baselines achieve much higher absolute success rates. ...

TL;DR for operators If a robotics team replaces a gripper, controller, sensor suite, or robot body, it should not automatically have to rebuild the system’s physical reasoning from scratch. Liang et al. argue that today’s embodied-AI stacks often make that reuse difficult because action semantics, coordinate frames, controller assumptions, verification logic, and model responsibilities remain entangled inside project-specific implementations.1 ...

TL;DR for operators A model can correctly recognize what is visible in a video and still lose track of who performed an action, how many times an event occurred, or which viewpoint observed it. GameplayQA1 makes those failures separately measurable. Across 16 evaluated multimodal models, average accuracy falls from 61.2% on single-reference tasks to 56.0% on temporal tasks and 49.4% on synchronized cross-video tasks. Occurrence counting averages just 36.5%, while cross-video ordering reaches 38.8%. Other-agent actions and states are also harder than questions about world objects. ...

TL;DR for operators A robot can perceive more of its environment than its planner should necessarily receive. In the experiments summarized here, adding task-irrelevant objects to structured scene representations increases the burden on classical planners and can leave harder problems unsolved. The proposed response is not a new end-to-end planner. It is a learned relevance layer that decides which objects and relations should survive into the planning problem. ...

TL;DR for operators A collaborative robot can execute its internal policy consistently and still behave unreliably from the human partner’s perspective. The reason is that the acceptable action may change with the person’s goal, gaze, gesture, language, movement, prior interaction, and surrounding context. Kwok and colleagues argue that reliability in human-robot collaboration should therefore be understood partly as maintaining sufficiently aligned interpretations between human and robot, rather than only as achieving predictable or robust model behaviour.1 ...

TL;DR for operators The most revealing moment in this study is not that GPT-5.1 moved a robot toward a plush penguin. After the robot struck the target, the model reversed to regain visual perspective, saw that the penguin was still upright, commanded another strike, then reversed again to verify the result. That sequence suggests something more operationally relevant than one-shot visual command generation: the controller retained a task state across several actions, interpreted the likely consequence of a collision, gathered new evidence, and corrected its plan. ...

TL;DR for operators A robotics system can generate a readable plan, revise that plan after an error, and improve its overall task-completion rate without demonstrating that its language component is correctly grounded in the physical environment. That attribution problem is the focus of a review by Yifan Guo and colleagues.1 The authors audit 105 foundation-model-enabled embodied-agent papers by separating two questions: what responsibility does language carry inside the system, and what evidence actually tests that responsibility? ...

TL;DR for operators A manipulation agent can reach the requested final state and still have executed the task unsafely. That creates a measurement problem for teams using task-completion rates to decide whether an embodied model, prompt, or policy update is ready for deployment. SafeRelBench tests this gap directly. Across seven evaluated VLM-driven agents, the spatial-relation cases produced task Success Rates (SR) of 0.52–0.73 but Safety Success Rates (SSR) of only 0.16–0.40. In matched non-spatial settings, SR rose to 0.83–0.94 and SSR reached as high as 0.91.1 The benchmark therefore measures something final-state success can miss: whether the agent satisfied the relevant safety prerequisite before taking the risky action. ...

TL;DR for operators Robotics teams often maintain separate models for navigation, arm manipulation, and hand interaction. That duplicates infrastructure and prevents each system from learning from the organization’s full pool of visual and physical experience. Worldscape-MoE1 shows that these controls can share one model without being forced through identical computation. Camera paths, robot commands, and hand-joint maps enter through pathways suited to their different structures, while the model activates shared computation alongside control-specific pathways. Under the reported shared training budget, this routed design outperforms dense mixed training in locomotion, manipulation, and hand-motion evaluation. The expected collapse from pooling unlike controls did not occur; performance weakened when every control had to use the same dense computation. ...