Research map / Learning theory and algorithms / Combinatorial and search optimization
Learning to delegate for large-scale vehicle routing
Sirui Li, Zhongxia Yan, Cathy Wu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| learned subproblem selector instead of Random selection (ablation) | iterative delegation with LKH-3 · Uniform CVRP N=2000 | speedup over LKH-3: 8.0x -> 15x | Improved |
| learned subproblem selector instead of Random selection (ablation) | iterative delegation with LKH-3 · Uniform CVRP N=500 | speedup over LKH-3: 3.8x -> 7.0x | Improved |
| learned subproblem selector instead of Count-based heuristic (ablation) | iterative delegation with LKH-3 · Uniform CVRP N=1000 | speedup over LKH-3: 8.2x -> 12x | Improved |
| learned subproblem selector instead of Max Min Distance heuristic (ablation) | iterative delegation with LKH-3 · Uniform CVRP N=2000 | speedup over LKH-3: 10x -> 15x | Improved |
| run Ours for twice the time (Long vs Short) | learned selector · Uniform CVRP N=2000 | cost: 234.89 -> 233.86 | Improved |
| iterative framework with learned selector instead of LKH-3 (30k) alone | LKH-3 · Uniform CVRP N=500 | cost: 61.87 -> 61.70 | Improved |
| hand-crafted selection heuristics in iterative framework instead of LKH-3 alone (ablation) | Random, Count-based, Max Min · Uniform CVRP | speedup over LKH-3: 5x to 10x | Improved |
| clustered CVRP distribution instead of uniform for N=2000 (Ours vs Random) | learned selector · Clustered CVRP N=2000, n_c=3 | speedup over LKH-3: 11x -> 26x | Improved |
| learned selector instead of Random | learned selector · Mixed CVRP N=2000, n_c=3 | speedup over LKH-3: 6.6x -> 13x | Improved |
| learned selector instead of Random | learned selector · CVRPTW N=2000 | speedup over LKH-3: 5.9x -> 8.2x | Improved |
| learned selector instead of Random | learned selector · VRPMPD N=2000 | speedup over LKH-3: 20x -> 31x | Improved |
| HGS subsolver instead of LKH-3, learned selector vs Random (ablation) | learned selector · Uniform CVRP N=2000 with HGS | speedup: 77x -> 103x | Improved |
Closest papers by meaning
- CaDA: Cross-Problem Routing Solver with Constraint-Aware Dual-Attention (ICML 2025 · similarity 0.89)
- Improving Generalization of Neural Combinatorial Optimization for Vehicle Routing Problems via Test-Time Projection Learning (NeurIPS 2025 · similarity 0.88)
- MVMoE: Multi-Task Vehicle Routing Solver with Mixture-of-Experts (ICML 2024 · similarity 0.88)
- Multi-Task Vehicle Routing Solver via Mixture of Specialized Experts under State-Decomposable MDP (NeurIPS 2025 · similarity 0.87)
- SHIELD: Multi-task Multi-distribution Vehicle Routing Solver with Sparsity and Hierarchy (ICML 2025 · similarity 0.87)
- Learning to Handle Complex Constraints for Vehicle Routing Problems (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
