Research map / Image generation / Fast diffusion sampling
Generative Large Neighborhood Search: Scalable Set Cover Optimization via Discrete Diffusion
Achref Jaziri, Thibaut Cuvelier, Bruno Adrien De Backer
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 |
|---|---|---|---|
| use denoising-trajectory sensitivity instead of final diffusion prediction only (ablation) | GLNS · Airline (OR-Lib) | primal gap (%): 9.52 -> 8.36 | Improved |
| use diffusion trajectory instead of static GNN embedding similarity (ablation) | GLNS · Rail unicost | primal gap (%): 4.41 -> 3.28 | Improved |
| use diffusion trajectory instead of spectral neighborhoods (ablation) | GLNS · Rail unicost | primal gap (%): 4.44 -> 3.28 | Improved |
| add random destroy-repair LNS as baseline vs learned RL policy | RL-LNS · Rail unicost | primal gap (%): 4.12 (Random-LNS) vs 4.5 (RL-LNS) | Hurt |
| replace constructive one-shot neural solvers with repair-based LNS | DIFUSCO · Rail unicost | primal gap (%): 6.42 -> 4.12 | Improved |
| use generative trajectory-guided selection over discriminative GNN-LNS | GNN-LNS · Airline crew scheduling | primal gap (%): 11.56 -> 9.1 | Improved |
| train on unicost, test on inverse costs: GLNS vs GNN-LNS | GLNS · Inverse (zero-shot) | primal gap (%): 2.98 -> 2.21 | Improved |
| use GLNS vs Gurobi with presolve under 60 s | GLNS · Large-scale Set Cover | primal integral: 995.57 -> 760.59 | Improved |
| enable Gurobi presolve | Gurobi · Large-scale Set Cover | primal integral: 1032.05 -> 995.57 | No clear effect |
| vary diffusion horizon, METIS variants, sub-solvers, cluster count (ablation) | GLNS · Set Cover | primal gap | No clear effect |
Closest papers by meaning
- Fast T2T: Optimization Consistency Speeds Up Diffusion-Based Training-to-Testing Solving for Combinatorial Optimization (NeurIPS 2024 · similarity 0.86)
- Regularized Large Neighborhood Search (NeurIPS 2026 · similarity 0.86)
- Guided Trajectory Generation with Diffusion Models for Offline Model-based Optimization (NeurIPS 2024 · similarity 0.86)
- Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization (NeurIPS 2026 · similarity 0.85)
- TFG: Unified Training-Free Guidance for Diffusion Models (NeurIPS 2024 · similarity 0.85)
- Large Language Model-driven Large Neighborhood Search for Large-Scale MILP Problems (ICML 2025 · similarity 0.85)
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.
