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

Abstract

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. Recent generative models, such as flow matching, have shown potential in Bayesian filtering. However, they still rely on particle-based representations of the posterior, which lose the rich information of the full distribution, or tackle a trajectory-level inverse problem that conflicts with the recursive structure of Bayesian filtering. To address this, we propose a new perspective of directly encoding the evolving distribution into flow matching model weights, namely, the Belief Flow Filter (BFF). It is a generative filtering framework that updates model weights via gradient descent at test time to track the posterior evolution. Thereby, BFF bypasses the scalability issue of particle representations or the flexibility limitation of Gaussian assumptions in conventional filters. We theoretically justify that the BFF design is structurally aligned with Bayesian filtering, and its training objective targets the recursive filtering operator. BFF is empirically verified across 5 different physical systems, including ones with chaotic dynamics and highly sparse, non-linear observations. The results show that BFF attains the best score in 8 of 9 metric-benchmark cells across the three standard 1D and 2D PDE benchmarks, and similarly leads on the extreme single-moving-sensor setting and on a real-world-grounded tokamak plasma estimation task, demonstrating its potential to accurately approximate the Bayesian filtering operator in high-dimensional probability space.

The authors’ abstract, as arXiv lists it. arXiv’s listings, abstracts included, are free for anyone to reuse.

More in Machine Learning (Statistics)

• arXiv preprint · 20 September 2026

  Stochastic Flow Map for Count Data

  Ganchao Wei
• arXiv preprint · 21 September 2026

  Exponential Family Synthetic Controls

  Hector Rodriguez-Deniz and David M. Blei
• arXiv preprint · 21 September 2026

  Doubly robust target inference for generalized linear regression with completely missing covariates

  Huali Zhao and Ke Deng
• arXiv preprint · 20 September 2026

  Distributional Balancing with Machine Learning for Clinical Trial Augmentation Using Real-World Data

  Zern Ke, Mingshi Cui, Gemma Moran and 1 other
• arXiv preprint · 21 September 2026

  Causal Bayesian Optimization: Foundations, Methods, and Applications

  Chenfeng Huang, Thuy T. Le, Zixuan Ma and 1 other

All new papers in Machine Learning (Statistics)