UNH at CheckThat! 2025: Fine-tuning Vs Prompting in Claim Extraction

Joe Wilder; Nikhil Kadapala; Benji Xu; Mohammed Alsaadi; Aiden Parsons; Mitchell Rogers; Palash Agarwal; Adam Hassick; Laura Dietz

UNH at CheckThat! 2025: Fine-tuning Vs Prompting in Claim Extraction

Computation and Language 2025-09-09 v1 Artificial Intelligence Information Retrieval

Authors: Joe Wilder , Nikhil Kadapala , Benji Xu , Mohammed Alsaadi , Aiden Parsons , Mitchell Rogers , Palash Agarwal , Adam Hassick , Laura Dietz

View on arXiv ↗ PDF ↗

Abstract

We participate in CheckThat! Task 2 English and explore various methods of prompting and in-context learning, including few-shot prompting and fine-tuning with different LLM families, with the goal of extracting check-worthy claims from social media passages. Our best METEOR score is achieved by fine-tuning a FLAN-T5 model. However, we observe that higher-quality claims can sometimes be extracted using other methods, even when their METEOR scores are lower.

Keywords

claim extraction check-worthy claims fine-tuning prompting in-context learning large language models

Cite

@article{arxiv.2509.06883,
  title  = {UNH at CheckThat! 2025: Fine-tuning Vs Prompting in Claim Extraction},
  author = {Joe Wilder and Nikhil Kadapala and Benji Xu and Mohammed Alsaadi and Aiden Parsons and Mitchell Rogers and Palash Agarwal and Adam Hassick and Laura Dietz},
  journal= {arXiv preprint arXiv:2509.06883},
  year   = {2025}
}

Comments

16 pages,3 tables, CLEF 2025 Working Notes, 9-12 September 2025, Madrid, Spain

UNH at CheckThat! 2025: Fine-tuning Vs Prompting in Claim Extraction

Abstract

Keywords

Cite

Comments

Related papers