Graph Algorithmic Reasoning on GAR Bipartite Matching
94.54AccuracyG-Substrate
Evaluation results
| Method | Links | |
|---|---|---|
| G-SubstrateTraining Paradigm=Multi-Task Training2026.01 | 94.54 | |
| Unified multi-taskTraining Paradigm=Multi-Task Training2026.01 | 94.23 | |
| GITATraining Paradigm=Task-Specific Training2026.01 | 93.19 | |
| Naive multi-taskTraining Paradigm=Multi-Task Training2026.01 | 92.21 | |
| G-WizTraining Paradigm=Task-Specific Training2026.01 | 92.15 | |
| Naive single-taskTraining Paradigm=Task-Specific Training2026.01 | 92.05 | |
| Naive multi-task + interleaveTraining Paradigm=Multi-Task Training2026.01 | 91.92 | |
| Unified single-taskTraining Paradigm=Task-Specific Training2026.01 | 85.98 |