Feature Extraction for Claim Check-Worthiness Prediction Tasks Using LLM

This study explores the use of Large Language Models (LLMs) for Claim Check-Worthiness Prediction (CCWP), a crucial pre-screening task in fact-checking. We predict the time between a claim’s occurrence and verification by analyzing data from fact-checking organizations. The results show that validation time is the same between the top 25% and bottom 75% of total checklist condition fulfillment claims. That is, further optimization is needed for LLMs to perform effective CCWPs.