The present four models were obtained by fine-tuning two base models on two different datasets for the lexical semantics task of hypernymy discovery.