GlucoPRISM (protocol-supervised CGM encoder; Trait/State/Sensor bands)
CGMOpen Weights
Key facts
- Authors
- Ahmed A. Metwally
- Pretraining Data
- 514 participants, 10,952 daily windows, 5 public CGM cohorts
Related Research
- BIG IDEAs Lab Glycemic Variability and Wearable Device Data
Research dataset
- GlucoPRISM: Protocol-Supervised Pretraining for Cross-Cohort Glucose Transfer
A CGM encoder built on the GlucoFM backbone whose 128-d representation is split by "protocol supervi
- GlucoFM
Foundation model
- OpenSLA
Sensor-Language-Action model (UCLA, Yuzhe Yang lab)