Prediction of Challenging Behaviors Associated with Profound Autism in a Classroom Setting Using Wearable Sensors

In three linesarXiv study using wearable sensors (accelerometry, EDA, temperature) and multimodal foundation models to predict challenging behaviors in 9 profoundly autistic children in classroom settings. Prediction up to 10 minutes in advance with AUC-ROC 0.78 on 110.7 hours of real-world data.

Summary generated by Claude — human-verified