Robots That Teach Themselves: Closing the Loop on Data Collection
A new system combines semantic planning and autonomous environment resets to enable robots to continuously gather training data without human intervention.
Science
Building AI with a Moral Compass
This review outlines a new framework for embedding human values into artificial intelligence, moving beyond abstract principles to concrete, verifiable standards.
When Swarms Meet: Steering Collective Behavior Through Impact
![Swarms of autonomous agents, each governed by target-velocity controls described by [latex]Eqs.(1)[/latex], demonstrate a capacity for cohesive redirection upon collision, transitioning from independent, fixed-formation travel to a transient milling phase before stabilizing into a unified composite with a resultant velocity distinct from the initial preferences of either constituent swarm.](https://arxiv.org/html/2603.12002v1/x1.png)
New research reveals how colliding groups of autonomous agents can be redirected, offering insights into controlling collective movement in multi-agent systems.
Mining GitHub for AI’s Building Blocks
A new framework automatically extracts reusable skills from open-source agent repositories, paving the way for more adaptable and capable AI systems.
Robots Learn by Imitation: Guiding Manipulation with Hand Shadows
Researchers demonstrate a vision-based system that allows robots to learn complex manipulation tasks by directly mirroring human hand movements.
Can AI Read the Science?
A new review explores the potential of artificial intelligence to automatically categorize biomedical research papers.
Robots That Don’t Fail: Ensuring Reliability in Learned Skills
As robots move beyond controlled environments, ensuring the dependability of learned behaviors is critical for safe and effective operation.
Rewriting Plasma Physics with Differentiable Programming
![Differentiable programming enables computational workflows that progress from manual iteration and costly parameter scans-scaling at [latex]\mathcal{O}(k^{N})[/latex]-to gradient-based optimization via reverse-mode automatic differentiation and, ultimately, function learning through neural networks embedded within differentiable solvers, representing the core focus of this work.](https://arxiv.org/html/2603.11231v1/x1.png)
A new approach leveraging automatic differentiation is transforming how we analyze, optimize, and design plasma systems.
Seeing is Doing: Robots Gain Smarter Perception for Complex Tasks
A new framework combines visual understanding and active perception, enabling robots to more effectively locate and manipulate objects in real-world environments.
Human Skill for Humanoids: Bridging the Dexterity Gap
Researchers have developed a new system that allows robots to learn complex manipulation tasks by directly leveraging human movement data.