adambuttrick/ror-set-point-matching-train-validation-test
Viewer • Updated • 434k • 125
Match affiliation strings to ROR records: dataset, retriever, set model, baseline and variants.
Note Data: all configs with train, validation and test splits
Note Retriever: finds candidate ROR records
Note Set model: selects the records that a string names
Note Baseline: pair scorer and chooser
Note Retriever variant: EmbeddingGemma
Note Retriever variant: no synthetic strings
Note Set model variant: bge-reranker-v2-m3
Note Set model variant: EmbeddingGemma