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

Sentence-level claim detection is a critical first step in the fact-checking process.While Large Language Models (LLMs) seem well-suited for claim detection, their computational cost poses challenges for real-world deployment.This paper investigates the effectiveness of both small and large pretrained Language Models for the task of claim detection.We conduct a comprehensive empirical evaluation using BERT, ModernBERT, RoBERTa, Llama, and ChatGPT-based models.Our results reveal that smaller models, when finetuned appropriately, can achieve competitive performance with significantly lower computational overhead on indomain tasks.Notably, we also find that BERTbased models transfer poorly on sentence-level claim detection in out-of-domain tasks, often over-predicting the positive outcome.We discuss the implications of these findings for practitioners and highlight directions for future research.