WearableQA tests AI reasoning on real users’ wearable data, a post says

The post says WearableQA uses hundreds of days of wearable measurements per person, supplemented by blood biomarkers and demographic information. It reports accuracy ranging from 19.6% to 72.9% across evaluated models, against a 10% chance baseline. Data reasoning was harder than health reasoning for most models, the summary says, while reasoning across multiple signals remained challenging across model scales. In the September 7, 2026 post, the user said the benchmark had not yet been released.