Reproduce “DEG: Diffusion-Enhanced GFlowNet for Solving Vehicle Routing 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

Ni Zhang, Zhiqin Zhang, Ling Pan, Hoong Chuin Lau, Zhiguang Cao

DEG combines edge-specific feedback from sampled routes, diffusion noise scheduling, and a scale adapter to extend GFlowNet routing to larger graphs. The noise gradually weakens edge guidance, while the adapter aligns test and training scales. On synthetic 10,000-node CVRP, DEG reports route length 1,125.04 and 24.95 seconds per instance, versus LEHD’s 1,213.13 and 556.88 seconds; no LKH comparison is reported.

Summary by alphaXiv.

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