Less Can be More: An Empirical Evaluation of Small and Large Language Models for Sentence-level Claim Detection FEVER WS