Entity Classification on RELBENCH rel-avito user-visits (test)
0.6701AUROCRel-LLM
Evaluation Results
| Method | Links | |
|---|---|---|
| Rel-LLMEvaluation Protocol=Fine-tuned2025.06 | 0.6701 | |
| FROG2026.05 | 0.6642 | |
| RDL2026.05 | 0.6635 | |
| RDLModel Type=GNN + Tabular2025.06 | 0.662 | |
| RelGNN2026.05 | 0.6567 | |
| ICL + MLPEvaluation Protocol=MLP on LLM representations2025.06 | 0.6498 | |
| GemmaTarget DB in pretraining=Maybe, Parameter count=27B2025.10 | 0.627 | |
| RTTarget DB in pretraining=Yes, Parameter count=22M2025.10 | 0.626 | |
| GriffinTarget DB in pretraining=Yes, Parameter count=22M2025.10 | 0.622 | |
| RTTarget DB in pretraining=No, Parameter count=22M2025.10 | 0.618 | |
| Entity MeanTarget DB in pretraining=No, Parameter count=02025.10 | 0.607 | |
| GriffinTarget DB in pretraining=No, Parameter count=22M2025.10 | 0.607 | |
| ICLEvaluation Protocol=In-context learning2025.06 | 0.6028 | |
| GemmaTarget DB in pretraining=Maybe, Parameter count=4B2025.10 | 0.601 | |
| GemmaTarget DB in pretraining=Maybe, Parameter count=12B2025.10 | 0.579 | |
| RelLLMTarget DB in pretraining=Yes, Parameter count=3B2025.10 | 0.562 | |
| Rel-ZeroEvaluation Protocol=Zero-shot2025.06 | 0.5617 | |
| LightGBMModel Type=Gradient Boosted Trees2025.06 | 0.5305 | |
| LightGBM2026.05 | 0.5282 |