🔬 AI Research

Academic papers, research breakthroughs, and scientific advances in machine learning and AI.

585 articles

AI detects hidden signs of solar storms nine hours before impact

A new AI system can spot subtle warning signs of incoming solar storms up to nine hours earlier than previous methods. Earlier detection could give power grids and satellite operators more time to prepare for disruptions. Source: universetoday.com

New multiple-instance learning method aims for broader use in pathology AI

Researchers propose a generalizable multiple-instance learning approach designed to work reliably across different computational pathology tasks. The method targets improved consistency when analyzing medical imaging data at scale. Source: nature.com

GC-MS and machine learning combine to detect bladder cancer markers in urine

Scientists used gas chromatography-mass spectrometry (GC-MS) together with machine learning to identify a chemical signature linked to bladder cancer in urine samples. The approach could pave the way for a non-invasive diagnostic tool. Source: chromatographytoday.com

Machine learning tackles 360° video streaming over 5G and beyond

A new research review examines how machine learning can improve the delivery of 360° video content across 5G and future B5G networks. It highlights emerging use cases as well as technical hurdles still facing widespread adoption. Source: link.springer.com

AI Model Predicts How Cancer Cells React to Different Drugs

Scientists are using data-driven models to forecast how various cancer cell types respond to different drug treatments. The approach aims to support more personalized and effective cancer therapies. Source: jacksonville.com

Machine Learning Cuts Half a Million Perovskites Down to 38 Solar Candidates

Researchers used a machine learning model to screen roughly 500,000 possible perovskite materials for solar cell applications. The approach narrowed the pool down to just 38 promising candidates for further study. Source: azom.com

Machine Learning Reveals Hidden Patterns in DNA Methylation

A new machine learning method has helped scientists detect previously unseen patterns in DNA methylation data. The findings could deepen understanding of how epigenetic changes influence gene activity. Source: phys.org

Scientists overcome complexity barriers in quantum learning

Researchers Youle Wang and Lei Zhang showed that physics-informed kernels can bypass long-standing query complexity limits in quantum learning tasks. The approach promises more efficient methods for learning quantum systems. Source: quantumzeitgeist.com

Machine learning could improve classification of genetic variants in prenatal testing

A new machine learning approach aims to identify disease-related genetic signatures in prenatal tissue by analyzing data from postnatal blood samples. The method could enhance accuracy in classifying uncertain genetic variants before birth. Source: progress.org.uk

Machine learning-driven enzyme evolution produces all versions of chiral alcohols

Researchers developed a stereodivergent protein engineering platform powered by machine learning that evolved a single alcohol dehydrogenase enzyme into eight variants. Each variant can produce a different stereoisomer of chiral alcohols, expanding options for precise chemical synthesis. Source: nature.com

Machine learning may predict antibiotic resistance in cystic fibrosis infections

A new study shows machine learning models can help forecast whether chronic lung infections in cystic fibrosis patients will resist commonly used antibiotics. This could support more targeted treatment decisions in the future. Source: cysticfibrosisnewstoday.com

Harvey Launches Tenet, a Kimi K3-Based Model for Legal AI Agents

Legal tech company Harvey has unveiled Harvey Tenet, a post-trained version of the Kimi K3 base model built with Fireworks for handling long, complex legal agent tasks. The company detailed its deployment approach, training methodology, and benchmark results. Source: marktechpost.com

AI Trained on Simulated Sites Reveals Hidden Features in Lidar Scans

Archaeologists have tested a new training method for deep learning models in cases where real examples of a given site type are rare. Instead of relying on large sets of known sites, the model was trained using simulated data to detect unknown features in lidar imagery. Source: archaeologymag.com

New AI Model Could Help Diagnose Peanut Allergy

A machine learning model has shown strong accuracy in telling peanut allergy apart from mere peanut sensitisation. Researchers say this could improve clinical diagnosis and reduce unnecessary testing. Source: emjreviews.com

Google Research Lets Environments Use Harness Framework Like AI Agents

A new Google paper presents a novel component that dynamically controls how complex operating environments interact with AI agents. The approach extends the Harness framework, previously used mainly for agents, to environments themselves. Source: eu.36kr.com

Random Survival Forests Outperform Classifiers in Colorectal Cancer Prognosis

A new study explores machine learning survival models for predicting colorectal cancer outcomes, showing random survival forests offer advantages over standard classification approaches. Colorectal cancer remains one of the leading causes of cancer death worldwide, driving demand for better prognostic tools. Source: nature.com

Georgia Southern University Adds Humanoid Robot to Engineering Program

Georgia Southern University has introduced an advanced humanoid robot within its Allen E. Paulson College of Engineering and Computing. The robot is meant to support both teaching and research activities for students. Source: griceconnect.com

Scientists Use Quantum Computer to Classify Neutrino Events

Researchers from UPV/EHU, Harvard, and TECNALIA have applied quantum computing to classify particle events recorded by neutrino telescopes. The findings were published in a quantum science journal. Source: quantumzeitgeist.com

AI Helps Scientists Crack the DNA Code Behind Gene Activation

Researchers have decoded the sequence of the 'initiator,' a key DNA element that triggers gene activation. Using machine learning analysis, the team identified patterns in this sequence that help explain how genes get switched on. Source: today.ucsd.edu

New Study Explores Scaling Laws for Data-Limited Language Model Training

As language models grow larger, their data requirements increase, but many valuable data sources, such as low-resource languages, remain scarce. A new study examines scaling laws for mixing different data sources under such constraints. Source: google.com