Reproduce “LyapuFlow: Controlling Generative Flows with Lyapunov Feedback for Inverse Problems” 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
Minseon Gwak, Hans Hao-Hsun Hsu, Danielle C. Maddix, N. Benjamin Erichson
LyapuFlow steers pretrained flows toward test-time constraints by predicting the terminal sample and measuring its constraint error. It applies feedback only when predicted error falls too slowly. On DIV2K motion deblurring, it improves SSIM and LPIPS over FlowDPS at 28 evaluations; on Burgers, it improves conditional fidelity over PCFM, which has zero constraint residuals while LyapuFlow’s are small but nonzero.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.