Reproduce “Learning to evolve by matching velocities” 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

Nicola Farenga, Andrea Manzoni

The authors introduce Velocity Matching (VM), a simulation-free method for learning continuous-time dynamics. VM estimates velocity between consecutive observations and trains the learned field at sampled points along their straight-line segment. Under regularity conditions, velocity error controls trajectory error. On the Stuart–Landau test, VM’s relative trajectory error was 1.89 × 10⁻² versus 4.36 × 10⁻¹ for teacher forcing; this benchmark does not establish universal superiority.

Summary by alphaXiv.

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