Research map / Graph learning / Graph self-supervised learning
Preference-Modulated Structural Attention for Multi-Objective Combinatorial Optimization
Rongsheng Jia, Jun Zhang, Yifan Zhang, Jian Cheng
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 preference modulation to edge structural bias (ablation) | PMSA · Bi-TSP100 | hypervolume: 0.7057 -> 0.7070 | Improved |
| add dynamic edge aggregation (ablation) | PMSA · Bi-TSP100 | hypervolume: 0.7053 -> 0.7070 | Improved |
| add preference modulation (ablation) | PMSA · Bi-TSP200 | hypervolume: 0.7384 -> 0.7408 | Improved |
| add dynamic edge aggregation (ablation) | PMSA · Bi-TSP200 | hypervolume: 0.7371 -> 0.7408 | Improved |
| add edge-feature structural bias vs WE-CA node-only model | PMSA · Bi-TSP100 | hypervolume: 0.7034 -> 0.7070 | Improved |
| use PMSA at small scale | PMSA · Bi-TSP20 | hypervolume: 0.6270 -> 0.6272 | No clear effect |
| use PMSA on larger unseen sizes | PMSA · Bi-TSP200 | hypervolume: 0.7346 -> 0.7408 | Improved |
| add instance augmentation | PMSA · Bi-TSP100 | hypervolume: 0.7070 -> 0.7083 | Improved |
| PMSA vs WE-CA on Bi-KP | PMSA · Bi-KP200 | hypervolume: 0.3602 vs 0.3602 | No clear effect |
| use PMSA on TSPLIB OOD instances | PMSA · KroAB100 | hypervolume: 0.6948 -> 0.6996 | Improved |
| PMSA-Aug vs POMO-Aug at balanced preference | PMSA-Aug · Bi-TSP100 lambda=(0.5,0.5) | gap: 0.77% -> 0.24% | Improved |
Closest papers by meaning
- Preference-Driven Multi-Objective Combinatorial Optimization with Conditional Computation (NeurIPS 2025 · similarity 0.87)
- Probing Neural Combinatorial Optimization Models (NeurIPS 2025 · similarity 0.86)
- BOPO: Neural Combinatorial Optimization via Best-anchored and Objective-guided Preference Optimization (ICML 2025 · similarity 0.86)
- $\texttt{STRCMP}$: Integrating Graph Structural Priors with Language Models for Combinatorial Optimization (NeurIPS 2025 · similarity 0.85)
- FERERO: A Flexible Framework for Preference-Guided Multi-Objective Learning (NeurIPS 2024 · similarity 0.84)
- Preference-Guided Diffusion for Multi-Objective Offline Optimization (NeurIPS 2025 · similarity 0.84)
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.
