Research map / Large language models / Agents, code and math
Bridging Synthetic and Real Routing Problems via LLM-Guided Instance Generation and Progressive Adaptation
Jianghan Zhu, Yaoxin Wu, Zhuoyi Lin, Zhengyuan Zhang, Haiyan Yin, Zhiguang Cao, Senthilnath Jayavelu, Xiaoli Li
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 progressive fine-tuning with LLM-evolved generators then real instances to LEHD | LEHD (RRC-50) · TSPLib 70 problems overall | optimality gap: 2.48% -> 1.05% | Improved |
| add progressive fine-tuning to LEHD on CVRPLib | LEHD (RRC-50) · CVRPLib SetX 100 problems overall | optimality gap: 5.90% -> 2.71% | Improved |
| add progressive fine-tuning to POMO on TSPLib large instances | POMO x8 aug · TSPLib [1000,5000) | optimality gap: 64.84% -> 32.43% | Improved |
| add Phase 1 (LLM-evolved synthetic generator fine-tuning) before TSPLib fine-tuning (ablation) | POMO x8 aug · TSPLib [200,1000) | optimality gap: 12.06% -> 6.78% | Improved |
| add Phase 1 before Phase 2 for POMO on large instances (ablation) | POMO x8 aug · TSPLib [1000,5000) | optimality gap: 44.85% -> 36.48% | Improved |
| add Phase 2 (fine-tune on real TSPLib instances) after Phase 1 for POMO (ablation) | POMO x8 aug · TSPLib [1000,5000) | optimality gap: 43.06% -> 36.48% | Improved |
| add Phase 2 for POMO on small instances (ablation) | POMO x8 aug · TSPLib [0,200) | optimality gap: 1.10% -> 1.11% | No clear effect |
| add Phase 1 for LEHD greedy on large instances (ablation) | LEHD greedy · TSPLib [1000,5000) | optimality gap: 8.7% -> 5.37% | Improved |
| add Phase 2 for LEHD greedy on medium instances (ablation) | LEHD greedy · TSPLib [200,1000) | optimality gap: 2.35% -> 2.39% | No clear effect |
| use LLM-evolved generator instead of manual seed generator (ablation) | POMO generator evolution (S1) · S1 validation set | avg aug gap: 19.79 -> 16.52 | Improved |
| add rank-based selection to generator evolution (ablation) | EvoReal evolution · S1 validation set | avg aug gap: 16.88 -> 16.52 | Improved |
| add design-guidance prompt to generator evolution (ablation) | EvoReal evolution · S1 validation set | avg aug gap: 16.62 -> 16.52 | Improved |
Closest papers by meaning
- Improving Generalization of Neural Combinatorial Optimization for Vehicle Routing Problems via Test-Time Projection Learning (NeurIPS 2025 · similarity 0.87)
- ReEvo: Large Language Models as Hyper-Heuristics with Reflective Evolution (NeurIPS 2024 · similarity 0.86)
- Neural Solver Selection for Combinatorial Optimization (ICML 2025 · similarity 0.86)
- GLOP: Learning Global Partition and Local Construction for Solving Large-Scale Routing Problems in Real-Time (AAAI 2024 · similarity 0.86)
- Interpreting Neural Combinatorial Optimization via Evolving Programmatic Bottlenecks (NeurIPS 2026 · 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.
