There are 307 papers in aging research that use foundation models: 5 trials and 75 observational studies in people, 2 in animals, 5 in cells, 9 reviews. The newest story, Nov 18, 2026: “GenVinci builds 3D tissue models from flat microscope images”.
The computational model predicted gene activity from tissue images and mapped states linked to senescence across 85.8 million cells.The computational model predicted gene activity from tissue images and mapped states linked to senescence across 85.8 million cells.
Combining two kinds of language analysis with thinking-test scores outperformed each approach alone across all eight tasks predicting genetic risk or brain protein…Combining two kinds of language analysis with thinking-test scores outperformed each approach alone across all eight tasks predicting genetic risk or brain protein buildup.
Who was studiedHebrew- and English-speaking participants, against each individual prediction approach
What changedCombining language measures and thinking-test scores improved prediction of genetic risk and brain amyloid across all eight tasks versus individual approaches
The catchGenetic risk and brain amyloid are stand-ins for Alzheimer’s diagnosis
An analysis of 3,103 scans from 1,653 participants found no consistent advantage over a conventional brain activity measure for assessing current cognition.An analysis of 3,103 scans from 1,653 participants found no consistent advantage over a conventional brain activity measure for assessing current cognition.
Who was studied1,653 participants from aging cohorts
What changedBrain scan model features showed no consistent advantage over conventional brain activity measures for assessing cognition or classifying diagnoses
The catchPerformance depended partly on feature extraction; different readouts improved selected cognitive associations
Lifestyle coaches preferred forecast-informed plans in 71.4% of paired reviews over plans based on wearable history alone.Lifestyle coaches preferred forecast-informed plans in 71.4% of paired reviews over plans based on wearable history alone.
Who was studied49 participants across behavioral personas, against Generic Coaching
What changedCertified lifestyle coaches preferred counterfactual-grounded plans for overall quality in 71.4% of evaluations against 28.6% for Generic Coaching
The catchCoaching plans were evaluated offline without following users to measure real-world habit adherence
In an observational study of 3,328 adults, a model using joint movements predicted 1,887 of 3,210 tested health measures.In an observational study of 3,328 adults, a model using joint movements predicted 1,887 of 3,210 tested health measures.
Median worst-group gaps in distinguishing outcomes were 0.028 by age, versus 0.004 by sex and 0.016 by race.Median worst-group gaps in distinguishing outcomes were 0.028 by age, versus 0.004 by sex and 0.016 by race.
Trained in a European cohort and tested in three others, the model estimated lifespan trajectories associated with observed mortality outcomes.Trained in a European cohort and tested in three others, the model estimated lifespan trajectories associated with observed mortality outcomes.
Who was studiedPeople across multiple geographic regions
What changedPersonalized lifespan trajectories modeled from retinal photographs were associated with observed mortality outcomes, including age of death
The catchHealth improvements and survival gains from using the model for screening were not measured