Research announcements from universities, hospitals and scientific institutions. Explore the stories, people and discoveries behind the headlines.
Dates and source material belong to the original release.
Read our research explainersBrowse the collection · original publication order
AI-evolved modular robots made of snap-together legs can run outdoors, flip upright when overturned, and keep moving even after losing limbs. Broken parts roll back and rejoin the group.
Read the releaseCerium magnesium hexalluminate displayed a continuum of states and no magnetic ordering -- hallmarks of a quantum spin liquid. But neutron scattering revealed the material achieves these effects through competition between ferromagnetic and antiferromagnetic interactions, not quantum mechanics.
Read the releaseWeill Cornell Medicine researchers built a machine learning model that provides rolling preeclampsia predictions during the third trimester using electronic health record data. Trained on 36,000 pregnancies and validated across three hospitals, the model was most accurate around 34 weeks, with blood pressure, lab results, and white blood cell count emerging as key predictors at different stages.
Read the releaseUCLA researchers developed a deep learning framework that generates three virtual stains simultaneously from autofluorescence images of unstained tissue. Board-certified pathologists confirmed the virtual stains matched traditional methods for identifying vascular invasion in thyroid cancer, potentially saving tissue, time, and cost.
Read the releaseResearchers at Penn State developed microscopic temperature sensors from two-dimensional materials that can be embedded directly onto computer chips. Each sensor is one square micrometer, responds in 100 nanoseconds, and exploits ion-electron coupling to deliver precise readings without draining chip performance.
Read the releaseAn analysis of 171 AI-generated images and 400 public comments across 17 charity campaigns found that when AI visuals are introduced, discussion shifts away from the cause: only 20% of comments engaged with the humanitarian issue, while 70% focused on AI ethics.
Read the releaseStanford materials engineers married century-old lead selenide semiconductors with modern gallium arsenide crystals to create bright infrared LEDs that tolerate billions of defects per square centimeter, potentially enabling cheaper infrared sensing for environmental and medical applications.
Read the releaseUniversity of Oregon physicists mathematically created the first ideal glass, a structure where amorphous molecules are packed as densely as possible. The material behaves mechanically like a crystal despite having no repeating structure.
Read the releaseResearchers at Xidian University have built photonic spiking neural network chips that perform both linear and nonlinear computation in the optical domain, achieving reinforcement learning with on-chip latency of just 320 trillionths of a second.
Read the releaseA Duke-NUS study in Nature Communications shows that swabbing air, cages, and water in live poultry markets detected 40 different viruses, including highly pathogenic H5N1 strains, even when direct bird testing found nothing. The approach is cheaper, safer, and more comprehensive.
Read the releaseAcoustic AI systems can identify mosquito species with up to 97% accuracy in lab conditions. But a study recording hundreds of wild mosquitoes in Hungary found that temperature, sex, and individual variation create real-world noise that current training data can't handle.
Read the releaseChildren's Hospital of Philadelphia researchers reviewed how generative AI affects children differently at ages 0-5, 6-11, and 12-plus, finding age-specific risks from blurred social boundaries in toddlers to unfiltered mental health responses for teens.
Read the releaseAn AI model trained on abdominal CT scans and paired clinical records outperformed specialist tools across more than 750 tasks - including predicting diabetes and heart disease onset years before symptoms appear. The work points toward a new class of general-purpose medical imaging AI.
Read the releaseWhen students in a mandatory data science course worked with examples from fields they cared about, their understanding and motivation improved measurably. A Japanese university offers a tested template for broader adoption.
Read the releaseFlorida Atlantic University's engineering college received a $4.5 million T-1A Jayhawk mixed-reality flight simulator from the U.S. Air Force Office of Scientific Research. The system offers a controlled, reconfigurable environment for research into AI decision-making, human performance, sensor fusion, and cybersecurity - at a fraction of the cost of live aircraft testing.
Read the releaseSimTac, developed by researchers at King's College London, is the first simulator capable of accurately modeling touch sensors with complex biological shapes. It combines particle physics, optical rendering, and neural networks to achieve real-time performance and near-perfect transfer from simulation to physical robots.
Read the releaseThe old way of finding better battery materials was to make them, test them, and see what happened. A comprehensive review in ENGINEERING Energy details how artificial intelligence is replacing that approach with one that starts from the desired outcome and works backward to the chemistry.
Read the releaseBefore quantum computers can be trusted with complex problems, engineers need to know exactly what operations they are performing - and whether those operations match the plan. A new framework from researchers in Japan and Vietnam makes that verification process far less expensive.
Read the releaseA research team led by Prof. Seunguk Song from the Department of Energy Science at Sungkyunkwan University (SKKU), in collaboration with the Institute for Basic Science (IBS), the University of Pennsylvania, and the U.S. Air Force Research Laboratory, has published a comprehensive technical roadmap for two-dimensional (2D) Indium Selenides (InSe)—a key material for next-generation low-power and quantum computing. The study, titled “Indium selenides for next-generation electronics and optoelectronics,” was recently published in Nature Reviews Electrical Engineering, the ...
Read the releaseA PLOS Biology analysis argues that AI models trained and tested on similar image datasets perform poorly when deployed in genuinely new ecological contexts - and often fail without signaling that failure to the user.
Read the releaseA team at the Shenzhen Institutes of Advanced Technology built a 3D facial database of roughly 200,000 high-fidelity scans and trained a curvature-based neural network on it. The resulting model identifies facial landmarks more robustly than previous methods that depended on 2D texture or synthetic faces.
Read the releaseSouthwest Research Institute has completed CMMC Level 2 certification through an independent third-party assessment organization, covering its Intelligent Systems and Mechanical Engineering divisions and satisfying 110 security requirements tied to protection of controlled unclassified information in government contracts.
Read the releaseMount Sinai and King Saud University Medical City announced a three-year research collaboration to study familial clustering of inflammatory bowel disease in Saudi Arabia, combining multi-omics analysis with structured biospecimen collection from families with multiple affected members to identify early diagnostic and therapeutic targets.
Read the releaseA controlled laboratory study comparing four small dogs (Chihuahuas) to three large dogs (mastiffs) found small dogs produced more airborne fine particles, likely from activity, while large dogs released more bacteria and fungi - including outdoor microorganisms not typically present from humans. The work appears in Environmental Science and Technology.
Read the releaseKorea Institute of Materials Science and Max Planck Institute researchers publish in Acta Materialia an explainable AI framework that predicts how internal defect morphology - not just porosity fraction - affects the mechanical properties of laser powder bed fusion parts, enabling defect-aware process design across steel, aluminum, and titanium alloys.
Read the release