Bayesian Filtering in Physical Systems via Test-time Trained Flow Matching

Summary

Bayesian filtering provides a principled framework for online state estimation under uncertainty, yet its application to systems with high-dimensional states and complicated posterior distributions remains challenging.