Reproduce “NAGO: Noise-Aware Generative Operator via Flow Matching” 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
Guanyu Chen, Shengze Xu, Pengwei Liu, Xingyu Ren, Pengkai Wang, Antonios Armaou, Tieyong Zeng, Dong Ni
NAGO designs source noise for functional flow matching to generate samples of fields. It selects an admissible Gaussian source whose spectrum matches the target over represented frequencies, aiming to reduce velocity-field stiffness. In a controlled 64×64 Matérn-field test, it reached spectral log-MSE 6.21×10⁻⁴ in five RK4 steps, versus about 30 steps for Gaussian-source flow matching; this result is specific to that setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.