CHAPTER Samo_LINGBUZZ_2025/IDIAP A structured synthetic dataset of English and Italian verb alternations for testing lexical abstraction via functional lexicon in LLMs. Samo, Giuseppe english italian LLM neural networks NLP semantics syntax verb alternations Lingbuzz / Archive paper 2025 https://ling.auf.net/lingbuzz/009085 https://ling.auf.net/lingbuzz/009085 URL In this paper, we develop a structured dataset the investigation of lexical abstraction via functional words in LLMs, concentrating on the change-of-state (COS) and object-drop (OD) verb alternations classes in English and Italian. We present the dataset creation process and report the results of baseline experiments.