Verifying Expertise: A New Layer of Biosecurity
A proposal suggests adapting financial ‘Know Your Customer’ protocols to govern access to powerful biological design tools and mitigate emerging biosecurity risks.
Science
When Robots See Actions That Aren’t There
New research explores why foundation models for robotics sometimes generate plans based on imagined events, and how to improve their reliability.
Beyond Plagiarism: Measuring true Originality in Science
A new framework uses the power of language models and existing research to move beyond simple copy-checking and assess the genuine novelty of scientific work.
Uncovering Stories of Cause and Effect
New tools are emerging to help researchers map relationships within qualitative data, moving beyond simple observation to explore underlying causal mechanisms.
Learning by Doing: How Actions Unlock Better AI Understanding
![The system considers how a physical entity, defined by a set of varying factors [latex]\mathbf{c}[/latex], responds to applied actions [latex]a\_{i} \in \mathbb{A}[/latex], where outcomes [latex]y\_{i}(\mathbf{c})[/latex] depend only on subsets of those factors, and demonstrates that any disentanglement of shared factors-like [latex]c\_{2}[/latex] required by multiple actions-will be achieved through a variational autoencoder architecture with separate encoders [latex]E\_{X}[/latex] and [latex]E\_{A}[/latex] processing input samples and action combinations to inform a shared decoder [latex]D[/latex] and ultimately predict system outputs.](https://arxiv.org/html/2602.06741v1/x1.png)
A new framework demonstrates that embedding actions within machine learning models improves their ability to learn and interpret complex systems.
Automating Particle Physics with Artificial Intelligence
A new framework harnesses the power of AI to streamline complex data analysis in high-energy physics, opening the field to a wider range of researchers.
Robots Learn by Watching: DreamDojo Builds a World From Human Video
A new foundation model allows robots to predict and interact with their environment by learning from vast datasets of human activity.
Building AI That Builds AI: The Rise of Agentic Software Engineers
As artificial intelligence systems take on increasingly complex development tasks, a new breed of AI – capable of acting as software engineers – is emerging, demanding a re-evaluation of what constitutes trustworthy AI.
Robots That Adapt: Navigating Uncertainty in Dynamic Environments
![Trajectory optimization under uncertainty-specifically with a half-width of 0.05m and variations in wall position between -0.7m and -0.3m, coupled with a restitution coefficient ranging from 0.7 to 0.9-was subjected to rigorous testing across 200 randomly sampled points within this parameter space, demonstrating the robustness of a five-branched SURE approach to nominal conditions of [latex]x_{\mathrm{wall}} = -0.5\,\mathrm{m}[/latex] and a restitution coefficient of 0.8.](https://arxiv.org/html/2602.06864v1/figures/robustness_comp_cond4.png)
A new framework enables robots to reliably interact with unpredictable environments by intelligently branching and merging potential trajectories.
Beyond the Hype: AI’s Role in Legal Fact-Finding
This review examines how generative AI is changing legal fact verification, focusing on the crucial balance between automation and maintaining professional expertise.