Research map / Label-efficient and robust learning / Knowledge distillation
Learning Generalizable Models for Vehicle Routing Problems via Knowledge Distillation
Jieyi Bi, Yining Ma, Jiahai Wang, Zhiguang Cao, Jinbiao Chen, Yuan Sun, Yeow Meng Chee
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 |
|---|---|---|---|
| add adaptive teacher selection (ablation) | AMDKD · CVRP-50 | avg gap %: 0.94 -> 0.90 | Improved |
| add KD loss (ablation) | AMDKD · CVRP-50 | avg gap %: 1.06 -> 0.90 | Improved |
| add task loss (ablation) | AMDKD · CVRP-50 | avg gap %: 1.00 -> 0.90 | Improved |
| use on-policy instead of off-policy distillation (ablation) | AMDKD · CVRP-50 | avg gap %: 10.42 -> 0.90 | Improved |
| use one teacher per epoch instead of multiple teachers simultaneously (ablation) | AMDKD-POMO · CVRP-50 | overall gap | Improved |
| distill AM with multi-distribution teachers into smaller student | AM · TSP-50 exemplar avg | gap %: 1.26 -> 0.91 | Improved |
| distill POMO into smaller student | POMO# · TSP-100 unseen OoD | gap %: 0.28 -> 0.41 | Hurt |
| distill POMO into smaller student | POMO# · CVRP-50 unseen OoD | gap %: 0.94 -> 0.90 | Improved |
| increase student dimension from 64 to 128 (ablation) | AMDKD-POMO · CVRP-50 | avg gap %: 0.90 -> 0.81 | Improved |
| start adaptive strategy at epoch 500 vs 1 (ablation) | AMDKD-AM · CVRP-50 | overall gap | No clear effect |
| increase validation set size from 1000 to 2000 or decrease to 500 (ablation) | AMDKD · CVRP | gap | No clear effect |
| change exemplar distributions to Expansion/Implosion/Explosion/Grid (ablation) | AMDKD-POMO* · CVRP-50 | avg gap %: 0.97 -> 0.92 | Improved |
Closest papers by meaning
- MTL-KD: Multi-Task Learning Via Knowledge Distillation for Generalizable Neural Vehicle Routing Solver (NeurIPS 2025 · similarity 0.91)
- Distilling Autoregressive Models to Obtain High-Performance Non-autoregressive Solvers for Vehicle Routing Problems with Faster Inference Speed (AAAI 2024 · similarity 0.90)
- Improving Generalization of Neural Combinatorial Optimization for Vehicle Routing Problems via Test-Time Projection Learning (NeurIPS 2025 · similarity 0.89)
- MVMoE: Multi-Task Vehicle Routing Solver with Mixture-of-Experts (ICML 2024 · similarity 0.89)
- Multi-Task Vehicle Routing Solver via Mixture of Specialized Experts under State-Decomposable MDP (NeurIPS 2025 · similarity 0.86)
- Memory-Enhanced Neural Solvers for Routing Problems (NeurIPS 2025 · similarity 0.86)
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.
