Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
Artificial intelligence models provide personalized advice, but may perpetuate negative stereotypes about people with autism. Researchers found that up to 70% of the time, AI discourages those with autism from socializing.
A team of scientists from NTU Singapore has developed a new biochip that, when paired with Artificial Intelligence (AI), can detect quickly and accurately extremely small amounts of microRNAs. The device can cut detection time from hours to 20 minutes.
A review article highlights a deep learning-driven CNN approach for detecting and classifying dynamic road obstacles, achieving high accuracy in obstacle identification and classification. The proposed architecture shows strong performance, but real-world deployment requires continued evaluation across larger and more varied scenarios.
A research team at Saarland University has developed an AI-assisted method to determine temperature distribution inside a running electric motor in real time, without additional hardware. The system uses motor-condition data extracted from electromagnetic fields and can detect thermal overload and optimize power regulation.
A team of researchers developed a machine learning framework to optimize laser settings for printing crack-susceptible superalloys. The algorithm reduced internal crack density by 99% and increased the metal's high-temperature strength, surpassing traditional cast components.
Researchers developed a new training technique, HarmonyGNN, to improve the accuracy of graph neural networks in heterophilic graphs. The framework achieved state-of-the-art performance on four heterophilic graphs with accuracy improvements ranging from 1.27% to 9.6%.
A novel imaging reconstruction framework, TT-PADM, enables high-quality photoacoustic tomography imaging even under limited-view and sparse-view acquisition constraints. This breakthrough technique reduces the number of required acoustic transducers without compromising image quality.
A custom-built AI system helped uncover how bacterial communities organize themselves, showing that early moments of a biological transition carry more information than previously considered. The findings bring new insight into the relationship between genotype and phenotype in Myxococcus xanthus.
A new study reveals that AI systems mimic the structure of human judgment but with a more rigid, rule-based approach. The researchers found biases in AI judgments, especially across demographic traits, highlighting the need for awareness and understanding how these systems 'think'.
The study uses adaptive machine learning force fields to track sodium metal-electrolyte reactions, achieving a 71% speedup over ab initio molecular dynamics while retaining comparable accuracy. The approach identifies key components of the solid electrolyte interphase, including Na2O and NaOH, which influence its stability.
Researchers identified patient-reported symptoms associated with GLP-1s, including menstrual changes, fatigue, and temperature-related complaints, that may not be fully captured in clinical trials or drug labeling. Nearly 4% of Reddit users reported reproductive symptoms, and fatigue was the second most common complaint.
Researchers from Tohoku University examined the role of materials databases in supporting modern artificial intelligence tools used in materials science. They found that database architecture can directly affect AI model performance and reliability. The study aims to improve database quality, connectivity, and develop new AI systems th...
Scientists at the University of Virginia Health System have developed a suite of AI-powered tools, called YuelDesign, YuelPocket and YuelBond, to transform how new drugs are created. These tools can design drug molecules tailored to fit their protein targets exactly, even accounting for protein flexibility.
A new machine-learning method detects sudden changes in fluid behavior, improving simulation capabilities for everyday applications like weather prediction and nuclear reactor safety. This enables faster design testing, real-time adjustments, and reduced computational burden.
Gladstone Institutes investigator Ryan Corces receives $750,000 to investigate unknown genetic variants contributing to Alzheimer's disease. He aims to identify new drivers and therapeutic targets using artificial intelligence and CRISPR tools.
Researchers developed Sandook, a software-based system that tackles three major sources of performance-hampering variability simultaneously. The two-tier architecture optimizes task distribution for the overall pool while faster schedulers on each SSD react to urgent events.
The Alliance for Clinical Trials in Oncology is enrolling adolescent and young adult cancer patients in various trials, including genetic services and treatment studies. These trials aim to address longstanding gaps in care and improve outcomes for AYAs with cancer.
NTU Singapore aims to integrate AI into 40% of its undergraduate courses by 2030. Half will use personalized learning, while the other half will teach students how to build and deploy AI agents. Students will have access to Google's premium AI tools and computing credits to create their own agents.
A new 3D printing preview tool, VisiPrint, uses AI to generate aesthetically accurate previews of fabricated objects, reducing the need for multiple reprints and waste. The system considers material properties, layer height, and nozzle path to create realistic simulations.
A new testing framework, Scalable Experimental Design for System-level Ethical Testing (SEED-SET), balances measurable outcomes and qualitative values like fairness. The system uses a large language model to capture stakeholder preferences and identifies scenarios where AI systems align with human values.
AI models exhibit emergent intelligence through specialization and cooperation, unlike physical systems where individual components reflect similar information, according to Bar-Ilan University research. This finding has implications for neuroscience and challenges traditional notions of intelligence in AI.
Researchers propose a foundational framework to help multi-agent, connected systems decide what information they can trust before acting. The 'cy-trust' concept assigns a numerical trust value between 0 and 1 to data from other agents based on sensing, context, network behavior, and past experience.
The Association for Computing Machinery has published its inaugural issue of ACM AI Letters, a premier venue for rapid and timely AI research. The journal aims to bridge the gap between traditional conferences and journals, featuring short, peer-reviewed contributions that accelerate knowledge dissemination across academia and industry.
A new class of ultra-high strength and ductility steel has been created using machine learning, achieving a rare balance of extreme strength and ductility. The resulting metal resists corrosion and degrades slowly in salt-water tests.
Binghui Wang, an Illinois Tech Assistant Professor of Computer Science, has been awarded the 2025 Distinguished Junior R&D Award by the IEEE Chicago Section for his groundbreaking research in AI security and trustworthy machine learning. The award recognizes Wang's impactful work on developing provably secure and trustworthy AI systems.
Current AI systems lack internal embodiment, a property that humans take for granted, which can lead to performance and behavior limitations. Researchers propose a dual-embodiment framework to guide future research in building safer and more aligned AI models.
A team of researchers at Binghamton University has developed a method to pinpoint discoveries that reshaped the course of science. The new metric uses neural embedding to analyze approximately 55 million scientific papers and patents, identifying major breakthroughs and simultaneous discoveries with greater accuracy.
A new AI chatbot helps patients access retinal detachment advice through personalized, real-time, clinically grounded conversations. The system outperformed leading large language models and includes accessibility features for people with low vision or limited English proficiency.
Researchers evaluated the performance of the Phoenix Sepsis Score for predicting in-hospital mortality among pediatric ICU patients in China. They found that a modified version, PSS+, showed substantially improved discrimination without sacrificing clinical usability.
Researchers have created a new AI model that can simulate molecules under extreme conditions, allowing for reliable discoveries in fields like drug development and sustainable chemistry. The model's stability opens up new opportunities for simulations in areas where long-term accuracy is essential.
Researchers have developed a reconfigurable textile interface that supports flat touch, folded 3D manipulation, and shape-change commands. The 'one cloth, many states' framework reduces the need to swap props and recalibrate alignment, making it promising for constrained-space operation training and human-machine interaction scenarios.
HeapGrasp uses RGB images to analyze object silhouettes and estimate its 3D shape, reducing the need for depth information. The approach achieves high accuracy while minimizing camera movement and execution time.
Researchers developed AI models to analyze conversations between children with cancer and their caregivers, detecting severe symptoms that require extra support. More complex prompting strategies outperformed simpler ones in accurately identifying symptom severity and its impact.
New research from the University of East London suggests that machine learning can improve the accuracy and nuance of personality tests like DISC assessment. Using over 1,000 participants, researchers achieved accuracy rates of over 93% in predicting personality types and identified four clear clusters with subtle overlaps.
Researchers have developed a new system to map gene expression across whole mouse bodies, providing a toolkit for studying molecular and cellular processes. The technique allows for the analysis of inflammation in every cell type and organ tissue, paving the way for a 'virtual mouse' model that could be used for research.
Researchers developed a machine learning model to predict 28-day mortality in patients with sepsis complicated by acute respiratory failure. The model demonstrated strong discrimination for predicting mortality, with key variables including oxygenation indices, serum albumin levels, and disease severity scores.
A Mayo Clinic study found that wearable sleep data can improve the prediction of patient engagement in a 12-week home pulmonary rehabilitation program. By combining baseline sleep data with machine learning and traditional clinical indicators, clinicians can tailor more effective care plans for patients with COPD.
A new deep learning model classifies Japanese Sue ware from 3D scans with high accuracy, using three-dimensional point clouds directly. The model achieved an overall accuracy of 93.2%, performing almost perfectly on visually distinct categories, while focusing on regions that may correspond to expert archaeologists' considerations.
Researchers developed an AI tool called PathogenFinder2 that can detect harmful bacteria before they infect humans. The tool uses protein language models and has been shown to significantly improve the detection of bacterial threats.
Researchers developed a machine learning model that analyzes patient demographics, electronic health record data, and blood test results to predict hepatocellular carcinoma (HCC) risk. The model achieved high accuracy and outperformed existing liver cancer risk prediction models, offering potential for widespread use in resource-limite...
Researchers developed an AI model that analyzes routine pathology slides to predict breast cancer recurrence and chemotherapy benefit. The model was validated in a large clinical trial and demonstrates fast, accessible, and globally scalable diagnostic capabilities.
The International Telecommunication Union (ITU) will host the seventh AI for Good Global Summit from 7 to 10 July 2026 at Geneva’s Palexpo convention centre. The summit aims to guide the future of artificial intelligence and unlock its potential to serve humanity.
Researchers from SDSU discovered surprising similarities among ancient writing systems from Africa and the Caucasus region. The study suggests the Armenian alphabet may be more closely related to the ancient Ethiopic writing system than previously thought, revealing possible cultural contact and influence between regions.
Researchers introduce the capabilities approach-contextual integrity (CA-CI) framework to address privacy and dignity risks in AI systems. The framework evaluates normative appropriateness of AI systems beyond narrow tasks and stable contexts, securing social life and human dignity.
Researchers at Dartmouth College developed a mathematical framework to map students' conceptual knowledge from short multiple-choice quizzes, revealing peaks of mastery and valleys of struggle. The technique could enable personalized learning, AI tutoring systems, and more efficient feedback.
A new AI pipeline called RAVEN has validated over 100 exoplanets, including 31 newly detected planets, using NASA's TESS data. The study found that around 9-10% of Sun-like stars host a close-in planet, and 'Neptunian desert' planets occur around just 0.08% of Sun-like stars.
A new study found that over 155,000 US deaths between March 2020 and December 2021 were not officially recorded as COVID-19 deaths, highlighting critical gaps in the death investigation system. These unrecognized deaths disproportionately affected certain populations, including racial and ethnic minorities.
Researchers have developed AI models to predict molecular electrostatic potentials, enabling rapid and accurate analysis of battery electrolytes. The study reveals that quadrupole moments provide valuable information for recovering the electrostatic landscape from simple point charges.
Researchers at TU Graz developed a neuroadaptive VR system that adjusts the intensity of exposure to anxiety levels, optimizing effectiveness in treating arachnophobia. The system uses EEG data and heart rate analysis, providing reliable indicators of stress and anxiety.
A team of computer scientists developed an algorithm that mimics bird flocking to help AI produce more reliable summaries of long documents. The framework reduces repetition and preserves key points, resulting in more accurate and concise summaries.
A new benchmarking study shows that AI coding tools struggle with structured outputs in software development, with even advanced models achieving only 75% accuracy. This highlights the need for human oversight and suggests that AI systems are not yet reliable enough to operate without human supervision.
Researchers developed a new multiview DNN structure to capture complex 3D anatomy and physiology from multiple imaging views, improving diagnostic accuracy for cardiovascular conditions. The approach demonstrated better performance than single-view DNNs and provided a viable alternative for other medical imaging modalities.
Researchers developed a computational tool that infers telomere length from structural changes in cells and tissues captured in medical biopsies. The TLPath model accurately predicts telomere length, providing new opportunities for studying human aging.
The new framework groups stations with similar hydrological behavior, reducing computational cost while maintaining high predictive accuracy. This approach enables scalable, data-efficient AI systems for water level forecasting, supporting flood early-warning systems, optimized reservoir and irrigation management, and improved decision...
The ARU Arm AI Lab will provide researchers and students with access to advanced Arm AI technology, focusing on real-world applications in healthcare and life sciences. This partnership will also support emerging talent and drive innovation, building on existing collaborations and industry projects.
A new photocatalyst design using machine learning interatomic potential calculations has successfully identified suitable dopants for a novel tin oxide material. The resulting aluminum-doped material produces 16 times more hydrogen under visible light than the undoped material, paving the way for next-generation clean energy applications.
Researchers developed predictive AI models that identified patients at risk of intimate partner violence from their medical records, years before they enrolled in care. The models achieved high accuracy and could detect IPV up to four years before patients sought care.
Kalinin's work is reshaping how new materials are designed, tested, and studied, enabling researchers to predict promising new materials computationally. He has developed machine learning-driven systems that can synthesize and characterize new materials at unprecedented speed.
Researchers developed an AI tool to predict patients at risk of intimate partner violence, using data from medical visits. The tool achieved high accuracy in detecting IPV among patients, with the multimodal fusion model outperforming others, detecting risk 88% of the time.
Researchers at MIT developed a new method that coaxes AI models to achieve better accuracy and clearer explanations in safety-critical applications. The approach extracts concepts the model has learned while training for a specific task and forces it to use those, producing better explanations than standard concept bottleneck models.