🔬 AI Research

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

585 articles

Apple's Minions turns idle Macs into a virtual GPU cloud to shorten AI queues

Apple researchers have published a paper describing Minions, a system that pools idle Apple Silicon Macs into a shared virtual GPU cloud for machine learning workloads. The aim is to cut waiting times for AI jobs by tapping unused computing capacity. Source: shattered.io

Neural networks can construct their own learning machinery, MIT study shows

MIT neuroscientists have demonstrated that an artificial neural network can build the mechanisms it needs to learn on its own. The network was given only simple instructions about how and when to learn, and it developed the rest itself. Source: news.mit.edu

UT San Antonio team builds model to sharpen machine-learning predictions

Researchers at UT San Antonio have developed a new model aimed at improving the accuracy of machine-learning predictions. Ngoc-Huynh Ho, PhD, an instructor in Radiology at the Long School of Medicine and the Biggs Institute, presented the work at a neuroepidemiology-focused conference. Source: news.uthscsa.edu

Machine learning extracts reaction kinetics from everyday yield data

University of Tokyo researchers created CYAN, a machine-learning method that pulls reaction rate constants out of the yield data chemists already gather while optimizing reactions. This means kinetic insight can be gained without dedicated, time-consuming kinetic experiments. Source: bioengineer.org

Machine learning guides new ceramic for smaller, temperature-stable capacitors

Researchers applied a hierarchical machine-learning strategy to identify tungsten bronze oxides and to steer bismuth and germanium substitutions. The approach led to a high-permittivity ceramic with improved properties for compact, temperature-stable capacitors. Source: azom.com

Transparent AI model forecasts student outcomes and explains its reasoning

Researchers built a hierarchical temporal graph network that predicts students' academic results with high accuracy. The model also relies on gradient-based techniques to reveal which factors drive its predictions, making its output interpretable. Source: bioengineer.org

Machine learning ties water quality data to possible pathogen risks

A research team created a framework that merges machine learning with quantitative microbial risk assessment. It aims to predict pathogen concentrations in water and estimate the associated health risks from routine quality measurements. Source: newswise.com

Hybrid AI hits 98% accuracy in power transformer fault prediction

Researchers developed a CNN-BiLSTM-Attention hybrid neural network tuned by an improved antlion optimization algorithm. The model diagnoses and predicts faults in power transformers, reaching 98% accuracy. Source: bioengineer.org

Sakana AI's LLM peer reviewer catches 73% of core-claim errors

Sakana AI released 'Beyond Imitation', a TMLR paper on LLM-assisted peer review centered on error detection. Unlike most AI reviewers, which are judged on how closely they mimic human reviews, this approach is evaluated on finding real mistakes. Source: marktechpost.com

Single accuracy scores mislead: ML models need distributional evaluation

A large-scale study argues that machine learning performance should be reported as full distributions, with confidence intervals on quantiles, rather than as one number. A lone accuracy score hides the variability that matters when comparing models. Source: bioengineer.org

AI turns 26 sediment samples into a map of an estuary's plastic legacy

Researchers built NIXVEGS, an open-source machine learning tool that predicts microplastic concentrations across an entire Baltic estuary from only 26 sediment samples. Source: bioengineer.org

Machine learning maps 10,000 reactions behind copper's CO2-to-fuel conversion

Researchers at the Indian Institute of Science built a machine learning framework that maps nearly 10,000 reactions in CO2 hydrogenation on copper, shedding light on how the metal turns CO2 into fuel. Source: bioengineer.org

Machine learning pinpoints microbial tipping points in bacterial vaginosis

A new machine learning study singles out ten key bacterial species and visualizes the abundance thresholds separating healthy vaginal microbiomes from dysbiotic ones. Source: bioengineer.org

Explainable ML predicts green concrete strength before the mix is poured

Researchers trained an explainable XGBoost model on 1,670 field-recorded concrete mixes and reached an R-squared of 0.9263 when predicting strength. SHAP analysis was used to show which factors drive the predictions. Source: bioengineer.org

AI shows how coating thickness shapes the roughness–pitting resistance link

Researchers at Shibaura Institute of Technology found a thickness-dependent interaction in steam-grown protective films on aluminum alloy, linking surface roughness to resistance against pitting corrosion. Source: eurekalert.org

Machine learning gauges corrosion and health risks in data-poor Algerian aquifers

Industrializing arid regions must make sure groundwater suits new industries while also protecting public health, often with very little data. A machine learning assessment tackles this dual challenge for aquifers in Northeastern Algeria, evaluating industrial corrosion and health risks. Source: nature.com

Machine learning links environment and immunity to childhood eczema risk

A machine learning analysis of 217 South African AmaXhosa children identified one protective and two susceptibility clusters connecting rural environment and immune factors to eczema. Source: bioengineer.org

Machine learning suggests chemoradiotherapy helps older head and neck cancer patients

A machine learning analysis of 1,525 SEER records finds that concurrent chemoradiotherapy is associated with roughly 30 percent better disease-specific survival in older patients with head and neck cancer. Source: bioengineer.org

Normalizing Trajectory Models Rethink Diffusion Sampling

Diffusion-based models break sampling into many small Gaussian denoising steps, an assumption that falters when generation has to be done in only a few steps. The work on Normalizing Trajectory Models targets this limitation of the standard approach. Source: machinelearning.apple.com

ETH Zurich and Roche Use Active Learning to Predict Where Molecules React

Researchers at ETH Zurich and Roche combined active learning with geometric and self-supervised deep learning to predict C–H borylation reactions. The approach aims to steer drug discovery by identifying which sites on a molecule are most likely to react. Source: bioengineer.org