Research applying artificial intelligence across science and engineering. Covers machine learning, sensing and quality control, immune and health prediction, and 5G networks.
The SNU team introduces Cluster-aware Upcycling, leveraging semantic structure of pretrained models to promote specialization among expert modules. This approach outperforms conventional Sparse Upcycling on image-text retrieval and various image classification benchmarks.
The new integrations provide enterprise AI agents direct access to live, structured research data from Dimensions' 430M+ interconnected records. This allows for AI-assisted analytics across one of the world's most comprehensive linked views of global research activity.
Assistant Professor Yingxue Zhang's project aims to develop urban AI models that can efficiently process vast amounts of human-generated data to optimize commute times, traffic safety, and more. The model will utilize offline reinforcement learning to tackle spatial-temporal dynamics in urban life.
Prof. Haim Sompolinsky receives the 2026 Dirac Medal for pioneering contributions to equilibrium statistical mechanics and non-equilibrium statistical mechanics. His work helps establish the field of theoretical neuroscience, elucidating how collective activity supports memory, computation, and learning.
New York's Empire AI Beta has officially launched, providing world-class AI computing power to researchers across the state. The initiative has served as a model for the federal National Science Foundation's State and Regional AI Infrastructure Hubs, which aim to build out regional AI research infrastructure and shared research capacity.
Recent studies examine the effectiveness of drones for prescription delivery and AI models in healthcare, raising questions about their impact on access and misinformation. Regulations around expertise and online content are also being reevaluated to address the root causes of these issues.
Non-invasive approaches are expanding options for assessing portal hypertension, with elastography techniques and biochemical markers showing high sensitivity and specificity. AI-powered predictive models combine clinical data to improve diagnosis, but should not replace invasive HVPG, which remains the gold standard.
A new study suggests that artificial intelligence can enhance the implementation of Medicaid work requirements by helping state agencies keep eligible individuals enrolled. AI tools can analyze existing databases to verify compliance or exemption status, reducing documentation difficulties and administrative complexities. However, huma...
Researchers developed a machine learning framework that predicts microbial contamination and estimates potential public health risks from routinely measured water quality indicators. The approach, called ML-QMRA, achieved high accuracy in predicting pathogen concentrations and their associated health risks.
UniSpec delivers lossless LLM acceleration without retraining while adapting automatically to different hardware platforms and multilingual workloads. The framework achieves up to 2.6× faster inference than existing methods across multiple models, hardware, and languages.
Researchers at Florida Atlantic University are developing a sustainable, AI-guided system to remove excess phosphorus from freshwater systems, preventing harmful algal blooms. The technology combines durable 3D-printed structures with AI-powered monitoring and optimization.
The Altmetric MCP brings real-world engagement data into AI workflows, capturing broader research impact beyond citation counts. This enables medical affairs teams and researchers to demonstrate the value of their work across societal platforms.
Researchers have found a new compound that promotes joint health and reduces inflammation-related genes in a model of osteoarthritis. The drug, M04, has shown promise as an innovative OA therapy by slowing down the disease process, giving people more years of pain-free living.
Assistant Professor Haonan Ling is exploring virtual solutions to biomanufacturing for the Genesis Mission, which aims to strengthen America's energy industry and national security. The project could significantly accelerate the development and deployment of sustainable biomanufacturing for fuels and chemicals.
The NFDI4Earth consortium facilitates access to environmental and climate data, while the DAPHNE4NFDI consortium preserves large-scale measurement datasets for long-term reuse. The NFDI-MatWerk and FAIRmat consortia develop common standards and digital tools for materials research, enabling faster and more targeted material development.
Researchers at Stanford University have developed an AI-powered tool called Evo 2 that can design novel phages to kill bacteria, including E. coli. The tool has already synthesized nearly 300 new phages and identified 16 exceptionally effective ones.
Researchers developed a programmable dynamic memtransistor that can process data at different speeds, reducing prediction errors by up to 40-fold. The technology enables accurate information processing even when input speeds vary.
New research suggests that collecting and spraying rainwater on rooftops during hot weather can significantly reduce energy demand for air conditioning and lower urban temperatures. The study found that this approach could cut the number of heatwave days overall and lessen the intensity of extreme heat events.
A new study found that six leading AI models exhibit a strong masculine bias in generating stories about animal characters, with only 2% being female. The models use neutral pronouns or avoid them altogether to reduce gender bias, but ultimately erase non-masculine identities.
A new AI triage agent will be embedded in Vanderbilt Health's EHR to quickly evaluate patients with Alzheimer's disease. The system will analyze chart information, flag missing details, and recommend referrals for priority or standard evaluation, aiming to reduce wait times from weeks to days.
A new AI model developed by Japan Advanced Institute of Science and Technology identifies the most relevant information in videos, significantly reducing analysis time. The model achieved state-of-the-art performance while using only 15.42% of available visual features and reduced processing time per video by approximately 65%.
Researchers developed GridFusionX, an AI tool that improves forecasts of electricity demand and renewable energy generation. The system reduces uncertainty in predictions, allowing grid operators to plan reserves more efficiently and respond proactively to sudden changes.
Jiaqi Ma's $660,307 grant aims to develop tools for understanding how individual components of training data affect large AI systems. This project will improve the performance and reliability of widely used technologies like language models and recommendation systems.
A WVU study found that ChatGPT-5 Pro can generate realistic psychiatry vignettes with strong diagnostic details but emphasizes the need for human-centered approach to ensure patient safety. The researchers recommend incorporating these vignettes into digital psychiatry curricula with faculty moderation and safeguards.
Researchers have developed a new photonic architecture that enables scalable spatiotemporal interleaving networks for high-density integrated photonic convolution. The SPIN (Spatiotemporal Photonic Interleaving Network) framework reduces waveguide complexity and increases programmability in wavelength-domain interleaving, enabling comp...
Ziyu Yao, Assistant Professor at George Mason University, has received a $674,100 NSF CAREER Award to advance AI interpretability research in code generation. She aims to develop a framework for mechanistic interpretation of language models and explore methods to improve code generation through interpretation-informed approaches.
Researchers at Harvard John A. Paulson School of Engineering and Applied Sciences developed an AI recommendation model that incorporates reinforcement learning to adjust to the uniqueness of each user. This approach improved human-AI performance over traditional one-size-fits-all decision support.
A qualitative study found social license for AI in healthcare is conditional and dynamic, influenced by structural, performance, and relational factors. The findings provide evidence-based recommendations for designing and implementing AI tools that support both consumers and clinicians.
Papers AI provides a fully integrated AI assistant that reads project drafts, datasets, and references together, reducing the need to re-explain context. The platform supports various file types, including Word, Markdown, Jupyter notebooks, and CSV files, and ensures data never leaves users' hardware.
The Smart QS Hackathon 2026 aims to explore innovative applications of Artificial Intelligence (AI) in Quantity Surveying (QS) through cross-sector collaboration, facilitating the intelligent transformation of the construction industry. Participants will have access to AI training sessions and workshops, as well as Cyberport's computin...
Massimiliano Albanese and Songqing Chen will host a two-day conference to strengthen secure and viable open-source ecosystems amidst the rise of artificial intelligence. The event aims to identify shared priorities and practical opportunities for improving long-term stewardship.
PaveX will develop an AI-sensor-based system to collect street-level data and produce practical road condition ratings, helping agencies plan repairs sooner and more efficiently. The project aims to improve distress detection reliability and reduce route planning overhead.
Professor Hengshuang Zhao has established a general framework for Physical Intelligence, spanning perception to understanding and virtual environments to physical reality. His work cites over 20,000 times and propels AI from perceiving to transforming the world.
Researchers from Stanford University found that using AI companions can exacerbate feelings of loneliness and isolation in vulnerable individuals. Participants who engaged with chatbots for emotional support reported lower well-being despite initial perceptions of improvement.
A research team from HKU School of Computing and Data Science has developed ClairS, a novel deep-learning algorithm that improves the detection of cancer mutations. Tested on breast, lung, and melanoma cell-line datasets, ClairS demonstrates exceptional accuracy across various cancer types and sequencing conditions.
A team from Tsinghua University developed the 'calculus of intelligence' framework to break down complex problems in agentic AI. This framework provides a mathematical approach to decomposing tasks into smaller, manageable subtasks that can be solved independently and then combined for a coherent whole.
The Center for Large Aperture Secure Sensing, Imaging and Communications aims to develop advanced antenna technologies for future wireless systems. Led by Edward Knightly, the research team will investigate how extremely large-scale antenna arrays can expand wireless system capabilities in challenging environments.
A team of researchers at UMass Amherst has developed an AI model, DiffuDose, that generates a patient's radiation dose map with gold-standard accuracy in under 23 seconds. This technology has the potential to unlock the full potential of radiopharmaceutical therapy for prostate cancer treatment.
NC State is leading three Genesis Mission projects using AI to accelerate discovery in high-performance computing, cybersecurity, and astrophysics. The projects focus on developing AI-powered tools for scientific research, defending agentic AI against adversarial attacks, and embedding AI techniques in large-scale simulations of neutro...
A team at Pusan National University created a stretchable organic electrochemical transistor that can be easily reprogrammed to perform different functions. The device combines logic, memory, and a visible color readout without complex circuitry.
Artificial intelligence can exacerbate or mitigate inequities in correctional health systems. Designing AI for access, inclusion, and trust is crucial for ensuring health equity among incarcerated populations. The study outlines policy and implementation considerations necessary to address these concerns.
The $24.9M grant will establish a national user facility for alloy-discovery campaigns using robotic systems and AI, freeing scientists to focus on discovery itself. ARM-MIP will democratize access to state-of-the-art research facilities and accelerate alloy development.
The University of Tennessee is expanding its research excellence with the addition of eight exceptional researchers, tackling pressing challenges in fields like precision health, advanced computing, and sustainable materials. The new faculty members bring expertise that builds on the university's strengths and solidifies its standing a...
A novel AI model can use information collected during routine sleep studies to identify patients' long-term health risks, including heart disease and cognitive decline. The findings also suggest that routine medical tests may contain substantially more physiologic information than current clinical practice extracts.
Researchers nationwide will have access to Georgia Tech's Advanced Manufacturing Pilot Facility remotely, leveraging AI, simulation, and autonomous experimentation to accelerate materials discovery. The cloud lab will expand who can take advantage of AMPF's capabilities, lowering costs and barriers to conducting research.
A new USC study uses AI to generate detailed maps of brain aging, revealing distinct patterns of neurodegeneration in specific regions. The approach sheds light on how local brain age correlates with changes in cognitive function across the lifespan.
A research team at KAIST has developed RL-SPH, a reinforcement-learning-based method that can independently generate feasible solutions satisfying all constraints. The technique achieved a 100% feasibility rate across five benchmarks, reducing the primal gap by an average of 28.6 times and improving search efficiency.
Recent studies explore the impact of digital technologies on health outcomes, highlighting innovations in maternal care and wildfire preparedness. Digital tools are being used to improve access to healthcare, particularly in rural areas, and mitigate adverse effects of wildfires on air quality.
The article highlights the need for AI tools to provide interpretable and reportable results, with a focus on improving turnaround time, diagnostic consistency, and clinical decision-making. While AI is unlikely to replace pathologists, it can help convert tissue morphology into actionable evidence for better patient care
Researchers at Seoul National University have developed a novel approach using porous triply periodic minimal surface (TPMS) feet and deep reinforcement learning controller, which significantly reduces battery power consumption in quadruped robots. The solution reduces energy consumption by up to 6.2% while maintaining stable locomotion.
Researchers introduce a new metric W² to complement traditional measures of predictive accuracy, revealing hidden patterns of systematic bias in AI models. By combining Q² with a penalty for angular difference from the ideal line, W² provides a more comprehensive evaluation of model reliability.
A new study by researchers at Johns Hopkins University highlights a significant disconnect between those who use AI health tools and those who create and fund them. The study reveals that key stakeholders have fundamentally different definitions of value, usability, and cost, creating systemic barriers to technology adoption.
A study by Bar-Ilan University researchers found that learning is driven primarily by changes in the strength of existing neural connections. The models became significantly better at learning as the amount of training data increased, but the proportion of lost connections remained roughly the same.
The collaboration enables researchers and innovators to develop, validate, and scale AI-enabled solutions across a broader representation of patient populations. This global health data network, called Mayo Clinic Platform_Connect, includes eight members and spans seven countries across three continents.
A KAIST research team developed two core technologies to correct AI hallucinations caused by sensory misinterpretation. The first technology uses the Diverse Negative Attributes method to accurately understand special camera sensors, while the second technology Modality-Adaptive Decoding blocks cross-modal hallucinations at the source.
Researchers developed a novel AI framework that optimizes investment decisions directly while accounting for risk. The study found that conventional forecasting-based approaches were outperformed by the decision-focused model in terms of risk-adjusted performance and wealth accumulation.
Scientists from the University of Osaka created an autonomous solid-state nanopore that can sense molecules, generate electrical signals, and retain memories of recent events. The device continuously changes its structure through chemical reactions, creating a dynamic sensing environment that responds to molecules passing through it.
A project led by Michela Taufer aims to accelerate AI-driven scientific discovery by enabling secure data sharing across the nation's research infrastructure. The National Science Data Fabric (NSDF) will connect researchers, computing resources, and data repositories, making advanced AI-driven science accessible to all.
Researchers at Duke University developed an AI framework called Raygun that can create modified versions of proteins while preserving structure and function. The tool uses protein language models to make extensive changes to existing proteins, opening up new avenues for engineering customized-sized proteins.
Researchers have created an AI model that accurately predicts cardiac index, a metric used to evaluate heart function, using non-invasive sensors on patient skin. The system demonstrates potential for accessible cardiovascular assessment beyond major hospitals and specialized clinics.