Seeing is Understanding: A Model That Thinks with Images
Researchers have developed a new AI model that dramatically improves scientific reasoning by actively manipulating images to test hypotheses and validate findings.
Science
Bridging the Gap: Robots Learn from High-Level Goals to Precise Actions
A new framework empowers robots to translate semantic intent into complex movements by decoupling high-level task understanding from low-level motor control.
Can AI Truly Understand Scientific Research?
A new benchmark assesses how well large language models can reason with and critique complex scientific papers, moving beyond simple information retrieval.
Decoding Robot Systems with AI: Can Language Models Understand ROS2?
This research explores whether large language models can effectively interpret the complex architectures of Robot Operating System 2 (ROS2) based robotic systems.
Unlocking New Physics with Neutrino Collisions
![Classification accuracy distinguishes between neutrino signal interference (NSI) data and that predicted by a [latex] \nu_e\nu_e [/latex]-coupled sterile model, with performance assessed across varying binning resolutions-defined by the number of divisions in length, energy, and time-and ranging from the threshold of random guessing (0.5) to perfect discrimination (1.0).](https://arxiv.org/html/2604.21869v1/x14.png)
Stopped-pion experiments, combined with advanced data analysis, offer a powerful new pathway to probe beyond the Standard Model.
Your Digital Shadow, Under Control
A new approach combines AI agents and user-defined privacy profiles to navigate the complex world of online consent and data management.
Universal Sensor Intelligence: A Model for Activity Recognition Across Diverse Environments
![The proposed human activity recognition model processes each data channel independently with a shared encoder, then integrates channel metadata via conditional batch normalization before fusing channel-wise features with mean pooling to generate a final prediction, all while simultaneously imposing auxiliary channel-specific predictions and optimizing performance through a combined loss function [latex]\mathcal{L}\_{\mathrm{comb}}[/latex] comprising both fused [latex]\mathcal{L}\_{\mathrm{fused}}[/latex] and channel-wise [latex]\mathcal{L}\_{\mathrm{dist}}[/latex] loss components.](https://arxiv.org/html/2604.21369v1/x2.png)
Researchers have developed a new framework for human activity recognition that overcomes the limitations of fixed sensor configurations, enabling a single model to adapt to a wide range of IoT devices.
Seeing is Manipulating: A Robot Learns From Every Angle
Researchers have developed a new framework enabling robots to reliably grasp and manipulate objects even when viewed from unfamiliar perspectives.
Beyond Search: An AI That Explores the Web to Generate Truly Novel Answers
Researchers have developed a new agentic AI system capable of independently exploring the web and synthesizing information in a way that surpasses traditional search-and-retrieve methods.
Robots Learn Faster with a Little Help from Humans
A new framework shifts robot training from the real world to a simulated environment, enabling scalable learning through interactive human correction.