We develop a variational approach to solve image inverse problems like super resolution, inpainting and deblurring for flow-based generative models.

This paper proposes a constrained posterior sampling approach for time series generation with hard constraints.

We determine the sample complexity of Bayesian recovery for solving inverse problems with general prior, forward operator and noise distributions.

We propose LFlow, a training-free framework for solving linear inverse problems using pretrained latent flow priors with theoretically grounded posterior guidance, achieving superior reconstruction quality over latent diffusion baselines.