Latent World Model for Kinematics
Latent World Model for Kinematics
Auto-generated stub. Edit this file to add more details.
A representation learning approach that uses predictive future-state objectives rather than raw signal reconstruction to capture structured kinematic dynamics in low-data regimes.
Why It Matters
Represents a shift from standard masked-reconstruction tasks to predictive future-state tasks in small-data inertial time series, offering a more robust representation for clinical downstream tasks.
Evidence
Sonata is a compact latent world model for six-axis trunk IMU representation learning under clinical data scarcity.
Related Papers
Metadata & Links
- created_at
- 2026-04-23T05:48:57Z