Ricerc@Sapienza

Roberto Navigli’s research focuses on Large Language Models (in particular with the creation of the Minerva LLM) - covering their pre-training, instruction tuning, safety, distillation, and parameter-efficient adaptation - together with the study of commonsense knowledge and reasoning capabilities. He also works extensively on multilingual natural language understanding, addressing tasks such as multilingual word sense disambiguation, entity linking, semantic role labeling, semantic parsing, summarization, machine translation and sentence representation. Anoter major part of his work concerns large-scale lexical-semantic knowledge acquisition, particularly the creation, curation, and maintenance of multilingual knowledge graphs and resources such as BabelNet.