Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.
By April 2024, researchers at UTM University's School of Education had tested ChatGPT acceptance for metacognitive self-regulated learning with 300 preservice teachers. Participants completed a scenario-based activity and a Technology Acceptance Model questionnaire, followed by reflections and interviews, with responses analyzed via structural equation modelling.
Published June 3, 2024 in Nature Communications, researchers described ChatMOF, an AI system built on GPT-4 and GPT-3.5 variants to handle metal-organic framework tasks from natural language. The pipeline includes an agent, toolkit, and evaluator to manage data retrieval, property prediction, and structure generation, reporting 96.9% searching, 95.7% predicting, and 87.5% generating accuracy with GPT-4.
Researchers surveyed 385 senior high school students in Ghana to test how teachers' AI-supported pedagogical practices relate to mathematics problem-solving, whether student engagement explains that link, and whether mathematical self-belief moderates it.
Published August 20, 2024, this peer-reviewed perspective examines multimodal large language models in medicine. It notes that clinical work depends on diverse data types from MRI and CT scans to EHR time-series, audio, text, video, and omics, while most current LLMs process only text.
By August 2026, a review in Clinical Microbiology and Infection summarized evidence on using unsupervised machine learning to derive clinical subphenotypes of bloodstream infections. The strongest data were in Staphylococcus aureus bacteraemia, where latent class and cluster analyses identified distinct subgroups with different mortality across cohorts, with emerging work in ICU mixed-pathogen and transplant populations and early bedside calculators for phenotype assignment.
Researchers applied a machine learning model called Mosamatic to routine CT staging scans from 615 patients with stage II-III colon cancer treated between 2015-2021 at one center to quantify skeletal muscle area, density, and fat. They found myosteatosis was more common in those who later recurred and was linked to worse recurrence-free survival at 1 and 5 years.
A review published August 26, 2026 describes echocardiography workflows transitioning from manual acquisition and measurement to AI-driven interpretation, citing prospective evaluations where AI reduces examination time, automates measurements, and supports integrated assessments of ejection fraction, myocardial texture, and Doppler hemodynamics.
The study used machine learning on naive CD4+ TCR and naive BCR AIRR-seq repertoires to test classification of celiac disease versus controls, building on prior work linking germline HLA variation to naive repertoire composition and earlier BCR-based classification attempts.
ace