Reproduce “Probabilistic Geodesic Flow Matching on Location-Scale Families” 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
Zeyuan Yu, Zhi Chang, Shiwei Lan
The authors introduce probabilistic geodesic flow matching, which routes distributions from noise to data by minimizing energy under the Fisher–Rao metric in an exponential-power family. On 100-dimensional Neal’s funnel, the q=1 hybrid path reports W₂ of 2.99 ± 0.238, versus 3.76 ± 0.306 for q=1 PG. Results are task- and metric-dependent, not evidence replacing optimal transport.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.