Research map / Image generation / Fast diffusion sampling
Efficient Few-Step Solution Generation via Discrete Flow Matching for Combinatorial Optimization
Yuanshu Li, Di Wang, Wei Du, Xuan Wu, Peng Zhao, Yubin Xiao, You Zhou
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 velocity-based consistency loss (ablation) | EFLOCO · TSP, 10-step inference | optimality gap (%): 0.14 -> 0.05 | Improved |
| add velocity-based consistency loss (ablation) | EFLOCO · TSP, 2-step inference | optimality gap (%): 0.19 -> 0.07 | Improved |
| use ATSS in training and inference instead of uniform training + linear (ablation) | DIFUSCO retrained · TSP, 10 steps | optimality gap (%): 0.5868 -> 0.4352 | Improved |
| use ATSS in training and inference instead of uniform training + linear (ablation) | DIFUSCO retrained · TSP, 20 steps | optimality gap (%): 0.5370 -> 0.3616 | Improved |
| use ATSS inference only (uniform training) vs cosine inference (ablation) | DIFUSCO retrained · TSP, 10 steps | optimality gap (%): 0.5471 -> 0.4743 | Improved |
| use EFLOCO instead of DIFUSCO (greedy) | NAR TSP solver · TSP-100 | optimality gap (%): 0.23 -> 0.38 | Hurt |
| use EFLOCO instead of MatNet (greedy) | ATSP solver · ATSP-50 | optimality gap (%): 1.13 -> 0.22 | Improved |
| use EFLOCO instead of T2TCO (greedy) | large TSP · TSP-500 | optimality gap (%): 0.91 -> 1.48 | Hurt |
| use EFLOCO instead of DIFUSCO (greedy) | large TSP · TSP-500 | optimality gap (%): 1.75 -> 1.48 | Improved |
Closest papers by meaning
- Generation as Search Operator for Test-Time Scaling of Diffusion-based Combinatorial Optimization (NeurIPS 2025 · similarity 0.88)
- Unsupervised Diffusion Solver for Combinatorial Optimization via Combinatorial Adjoint Matching (ICML 2026 · similarity 0.88)
- Boosting Cross-problem Generalization in Diffusion-Based Neural Combinatorial Solver via Inference Time Adaptation (AAAI 2026 · similarity 0.88)
- Fast T2T: Optimization Consistency Speeds Up Diffusion-Based Training-to-Testing Solving for Combinatorial Optimization (NeurIPS 2024 · similarity 0.87)
- Large Language Models as End-to-end Combinatorial Optimization Solvers (NeurIPS 2025 · similarity 0.85)
- Efficient Combinatorial Optimization via Heat Diffusion (NeurIPS 2024 · 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.
