Reproduce “RW-Flow: One-Step Generation on Compact Manifolds via Wasserstein Gradient Flows” 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
Ualibyek Nurgulan, Seungwoo Yoo, Prin Phunyaphibarn, Minhyuk Sung
RW-Flow extends W-Flow to compact manifolds by learning a direct map from a starting point to a generated point. Sinkhorn couplings train the sampler, which needs one evaluation. The authors prove identifiability when the associated Gibbs kernel distinguishes signed measures. On Bunny mesh sampling, times are 0.16 versus 49.5 seconds; quality measures are mixed, and this is not universal.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.