Reproduce “Normalizing Flow Maps” 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

Francesco M. Ruscio, Timothy James Hitge, Fred Zhangzhi Peng, Danyal Rehman, Zander W. Blasingame, Nicholas M. Boffi, Michael M. Bronstein, Alexander Tong and 1 more

Normalizing Flow Maps (NFMs) learn invertible jumps along a noise-to-data path, enabling few-step generation and exact endpoint likelihoods through the change-of-variables rule. On class-conditional ImageNet-64, NFM-D uses 123M parameters and reports FID 4.59 and FLD 6.78, versus 122M, 4.89, and 6.98 for iTARFlow. NFM-R has stronger image results but may lose numerical invertibility.

Summary by alphaXiv.

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