Building Reliable Scientific Agents with Structured Execution
A new framework leverages typed execution graphs to enhance the reproducibility and scalability of complex scientific workflows powered by large language model agents.
Science
Beyond ‘AI’: Why Precision Matters
Overbroad discussions of ‘artificial intelligence’ obscure critical details and impede effective oversight, particularly when considering its application in sensitive areas.
Soft Robotics Takes a Leap Forward
Researchers have developed a novel soft robotic actuator and quadruped capable of dynamically switching between walking and jumping gaits.
The Rise of AI Evaluation: Measuring the Next Generation of Intelligent Systems
As artificial intelligence moves beyond isolated models to complex, autonomous agents, rigorous evaluation is no longer a supporting task, but a critical necessity.
Smarter Science Chatbots: How to Build Them Right
New research evaluates the best methods for equipping AI chatbots with the knowledge they need to answer questions about complex scientific literature.
AI Code Assistants and the Rise of TypeScript’s Hidden Debt
New research reveals that while AI agents are quickly adopted for code contributions, they often introduce type-related issues in TypeScript projects.
Less is More: A Minimalist Approach to Robot Learning
Researchers have demonstrated that a surprisingly simple vision-language-action model can achieve state-of-the-art performance in robotic manipulation tasks.
Letting Language Models Discover Science
![ModelSMC automatically discovers models from textual problem formulations and context data by iteratively refining an initial model-inspired by Sequential Monte Carlo methods that approximate distributions via weighted particles-to sample high-density regions of the model posterior [latex] p(m|\bm{x}_o) [/latex], effectively approaching the underlying data-generating process through model propagation via large language model sampling and weighting based on likelihood evaluation.](https://arxiv.org/html/2602.18266v1/x1.png)
A new framework harnesses the power of large language models to automate the process of scientific model discovery, moving beyond traditional methods.
Seeing is Understanding: Robots Learn to Interact with the World Through Language
![The system iteratively refines its understanding of a scene through interactive perception, where a robot evaluates observations and queries-leveraging a memory [latex]\mathcal{M}_{t}[/latex] to avoid repetition-and, when necessary, annotates images with segmented elements like push lines, keypoints, or grid patterns to inform subsequent actions [latex]a_{t}=\pi(\mathcal{M}_{t},\textbf{x}_{t},z_{t},\tilde{o}_{t})[/latex], ultimately building a contextual history [latex]S_{t}[/latex] of images and scene descriptions to enhance task efficiency.](https://arxiv.org/html/2602.18374v1/Figures/rss25/InterPer2.png)
A new framework empowers robots to manipulate objects and answer questions about their environment, even when vision is limited, by combining language understanding with enhanced visual perception.
Who’s Responsible When AI Fails?
A new framework seeks to pinpoint human accountability for the actions of increasingly complex artificial intelligence systems.