Entity Classification on RELBENCH rel-avito user-clicks (test)
71.31AUROCICL + MLP
Evaluation Results
| Method | Links | |
|---|---|---|
| ICL + MLPEvaluation Protocol=MLP on LLM representations2025.06 | 71.31 | |
| GelGT2026.05 | 68 | |
| FROG2026.05 | 67.22 | |
| Rel-LLMEvaluation Protocol=Fine-tuned2025.06 | 66.74 | |
| RelGNN2026.05 | 66.31 | |
| Rel-LLM2026.05 | 66 | |
| Nodeformer2026.05 | 66 | |
| RDLModel Type=GNN + Tabular2025.06 | 65.9 | |
| RDL2026.05 | 64.87 | |
| RelLLMTarget DB in pretraining=Yes, Parameter count=3B2025.10 | 62.3 | |
| Rel-ZeroEvaluation Protocol=Zero-shot2025.06 | 62.28 | |
| ICLEvaluation Protocol=In-context learning2025.06 | 61.32 | |
| RTTarget DB in pretraining=Yes, Parameter count=22M2025.10 | 60.9 | |
| GemmaTarget DB in pretraining=Maybe, Parameter count=27B2025.10 | 59.8 | |
| GemmaTarget DB in pretraining=Maybe, Parameter count=12B2025.10 | 59.5 | |
| RTTarget DB in pretraining=No, Parameter count=22M2025.10 | 59.5 | |
| LightGBM2026.05 | 54.65 | |
| GemmaTarget DB in pretraining=Maybe, Parameter count=4B2025.10 | 54.5 | |
| LightGBMModel Type=Gradient Boosted Trees2025.06 | 53.6 | |
| GriffinTarget DB in pretraining=No, Parameter count=22M2025.10 | 45.9 | |
| GriffinTarget DB in pretraining=Yes, Parameter count=22M2025.10 | 45.9 | |
| Entity MeanTarget DB in pretraining=No, Parameter count=02025.10 | 44.7 |