Reproduce “AdjointFlow: Accelerating Scientific Inverse Problems with Physics-Based Transport” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Ethan Hanold, Nicholas M. Boffi, Qin Li, Leonardo Zepeda-Núñez

AdjointFlow reconstructs wave inverse-scattering images from a physics-informed approximate inverse, learning a refinement path with target prediction and noise-free velocity. On Brain Scans, it achieves 3.421% RRMSE at eight sampling steps, versus 34.990% for a same-architecture conditional diffusion model. This equal-step-count comparison is not equal-time; accuracy matches or improves on five of seven settings.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.