Reproduce “Size Generalization for Diffusion Routing Solvers” 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
Barak Itzchak Davidovitch, Eli Meirom, Ethan Fetaya, Gal Chechik
ResizeCO adapts a diffusion solver supervised on labeled small instances to larger unlabeled routing problems using route-quality feedback, then improves routes with parallel GPU search. In synthetic CVRP-1000-to-10000, it reports a −0.096% average relative excess-cost gap against a heuristic reference in 49.9 minutes total, versus MaskCO’s 2.36% in 10.6 hours. The negative gap is not proof of optimality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.