We propose a synergistic two-stage framework that learns to jointly reconstruct physical fields and to optimize sensor placements with theoretical guarantees, both aiming for accurate physics sensing.

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

We develop flow map methods for state-of-the-art few-step generation, generalizing flow, diffusion, and consistency models.

EraseFlow surgically removes unwanted concepts with reward-free alignment, preserving prior and setting a new standard for safe, efficient diffusion models.