Centrala begrepp
Large Language Models with retrieval augmentation enhance fake news detection by strategically extracting evidence from the web.
Sammanfattning
The proliferation of fake news has significant implications on society. Traditional methods rely on outdated data, while Large Language Models (LLMs) offer a new approach. A novel framework, STEEL, combines LLMs and strategic internet-based evidence retrieval. The model ensures comprehensive evidence acquisition through multi-round retrieval strategies. Experiments across real-world datasets validate the framework's superiority over existing methods. The model not only provides accurate verdicts but also offers human-readable explanations for result interpretability.
Statistik
In November 2004, voters statewide voted to allow Miami-Dade and Broward counties to put questions on the ballot about adding slot machines at facilities.
Slot machines in Miami-Dade and Broward counties have generated 20 percent of the promised $500 million per year for schools.
One in every five families in New Jersey has a loved one with a mental illness.
Statistics of datasets: #Real News - 9,252; #Fake News - 3,555; #Total - 12,807.
Citat
"The model not only provides accurate verdicts but also offers human-readable explanations for result interpretability."
"STEEL outperforms state-of-the-art methods in both prediction and interpretability."
"Our work is a preliminary attempt to address systemic risks in the field of fake news detection."