List your company
ResearchFormerly WizWand

Entity Classification on RELBENCH rel-avito user-visits (test)

0.6701AUROCRel-LLM
0.5225240.5608370.599150.637463Jun 6, 2025Aug 3, 2025Sep 30, 2025Nov 27, 2025Jan 24, 2026Mar 23, 2026May 20, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Rel-LLMEvaluation Protocol=Fine-tuned2025.060.6701
FROG2026.050.6642
RDL2026.050.6635
RDLModel Type=GNN + Tabular2025.060.662
RelGNN2026.050.6567
ICL + MLPEvaluation Protocol=MLP on LLM representations2025.060.6498
GemmaTarget DB in pretraining=Maybe, Parameter count=27B2025.100.627
RTTarget DB in pretraining=Yes, Parameter count=22M2025.100.626
GriffinTarget DB in pretraining=Yes, Parameter count=22M2025.100.622
RTTarget DB in pretraining=No, Parameter count=22M2025.100.618
Entity MeanTarget DB in pretraining=No, Parameter count=02025.100.607
GriffinTarget DB in pretraining=No, Parameter count=22M2025.100.607
ICLEvaluation Protocol=In-context learning2025.060.6028
GemmaTarget DB in pretraining=Maybe, Parameter count=4B2025.100.601
GemmaTarget DB in pretraining=Maybe, Parameter count=12B2025.100.579
RelLLMTarget DB in pretraining=Yes, Parameter count=3B2025.100.562
Rel-ZeroEvaluation Protocol=Zero-shot2025.060.5617
LightGBMModel Type=Gradient Boosted Trees2025.060.5305
LightGBM2026.050.5282