Graph reasoning on GraphQA (test)
98.8Avg AccuracyGTLM
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GTLMrepresentation=Incidence2026.05 | 98.8 | — | — | — | — | — | — | — | — | — | — | — | — | |
| GTLMrepresentation=Standard2026.05 | 90.7 | — | — | — | — | — | — | — | — | — | — | — | — | |
| AutoBRANE2025.11 | 88.7 | 99.6 | 99.7 | 100 | 73.1 | 100 | 99.8 | 94.6 | 98.2 | 90.4 | 47.7 | 61.8 | 99.1 | |
| GradTAGapproach=Multi-Task Learning baseline2025.11 | 86.7 | 100 | 99.2 | 100 | 68 | 99.9 | 99.4 | 81.2 | 97.7 | 91 | 46.9 | 62.1 | 94.9 | |
| MTNapproach=fine-tuning a single adapter2025.11 | 85.9 | 96.7 | 98.1 | 100 | 67.2 | 99.5 | 99.4 | 81.2 | 98 | 86.6 | 48.1 | 60.8 | 94.9 | |
| 1st GraphTokenencoder=GCN2026.05 | 26.4 | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2nd GraphTokenencoder=MPNN2026.05 | 25 | — | — | — | — | — | — | — | — | — | — | — | — | |
| 3rd GraphTokenencoder=MHA2026.05 | 24.4 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Zero-Shot LLMmode=zero-shot2026.05 | 14.7 | — | — | — | — | — | — | — | — | — | — | — | — |