Learning Human Health and Diseases from 24-hour Wrist Movement "Here, we present Sensori, a self-supervised foundation model that learns general-purpose health representations directly from 24 hours of raw tri-axial wrist movement." Dataset: 122,640 participants contributing 683,617 person-days of free-living recordings. Architecture: Sensori uses a multiscale architecture in which pooling operations progressively reduce the temporal resolution. Training: Sensori is pretrained using two complementary objectives designed to capture movement patterns at different temporal scales: masked reconstruction and day-level contrastive learning. Results: adding Sensori embeddings to the clinical covariate model significantly improved AUROC for 52 of 102 eligible conditions across the six disease categories, with the largest gains for neurological and psychiatric disorders. paper link: arxiv.org/abs/2608.29494