Offline Multi-Objective Optimization on Off-MOO-Bench MORL
1.2Average IGDoffline RankD(best)
Evaluation results
| Method | Links | |
|---|---|---|
| D(best)Number of solutions=256, Percentile=50th2026.06 | 1.2 | |
| Multiple ModelsExtension=RoMA2026.06 | 2.6 | |
| DOMOONumber of solutions=256, Percentile=50th2026.06 | 3.1 | |
| End-to-End2026.06 | 4.3 | |
| Multiple Models + IOMNumber of solutions=256, Percentile=50th2026.06 | 4.6 | |
| Multi HeadNumber of solutions=256, Percentile=50th2026.06 | 4.9 | |
| ParetoFlow2026.06 | 5 | |
| Multiple ModelsExtension=COMs2026.06 | 5.4 | |
| MultipleModels + COMsNumber of solutions=256, Percentile=50th2026.06 | 6 | |
| Multiple Models2026.06 | 6.1 | |
| Multi Head + PcGradNumber of solutions=256, Percentile=50th2026.06 | 6.2 | |
| DOMOO2026.06 | 6.7 | |
| D(best)2026.06 | 6.8 | |
| Multiple Models + RoMANumber of solutions=256, Percentile=50th2026.06 | 7.2 | |
| Multi Head2026.06 | 7.5 | |
| Multiple ModelsExtension=IOM2026.06 | 7.5 | |
| Multiple ModelsExtension=ICT2026.06 | 7.7 | |
| Multiple Models + ICTNumber of solutions=256, Percentile=50th2026.06 | 8 | |
| Multiple Models + Tri-MentoringNumber of solutions=256, Percentile=50th2026.06 | 8.4 | |
| End-to-End + PcGradNumber of solutions=256, Percentile=50th2026.06 | 8.9 | |
| Multiple ModelsExtension=Tri-Mentoring2026.06 | 8.9 | |
| Multi HeadOptimization=PcGrad2026.06 | 9.4 | |
| End-to-EndOptimization=PcGrad2026.06 | 10.2 | |
| Multiple ModelsNumber of solutions=256, Percentile=50th2026.06 | 10.4 | |
| End-to-EndNumber of solutions=256, Percentile=50th2026.06 | 10.5 | |
| ParetoFlowNumber of solutions=256, Percentile=50th2026.06 | 10.83 | |
| End-to-End + GradNormNumber of solutions=256, Percentile=50th2026.06 | 11.4 | |
| Multi HeadOptimization=GradNorm2026.06 | 12.1 | |
| End-to-EndOptimization=GradNorm2026.06 | 12.3 | |
| Multi Head + GradNormNumber of solutions=256, Percentile=50th2026.06 | 13 |