Clickbait Detection via Large Language Models

TL;DR

This paper investigates whether large language models can serve as high-quality clickbait detectors. It systematically evaluates few-shot and zero-shot performance of GPT-3.5, GPT-4 (ChatGPT), and other baselines across English and Chinese datasets. The findings show LLMs generally underperform compared with fine-tuned PLMs, although GPT-4 offers improvements in comprehension and multilingual capability. The work highlights the need for methods that effectively harness LLMs for robust, language-agnostic clickbait detection and provides a public repo with results.

Abstract

Clickbait, which aims to induce users with some surprising and even thrilling headlines for increasing click-through rates, permeates almost all online content publishers, such as news portals and social media. Recently, Large Language Models (LLMs) have emerged as a powerful instrument and achieved tremendous success in a series of NLP downstream tasks. However, it is not yet known whether LLMs can be served as a high-quality clickbait detection system. In this paper, we analyze the performance of LLMs in the few-shot and zero-shot scenarios on several English and Chinese benchmark datasets. Experimental results show that LLMs cannot achieve the best results compared to the state-of-the-art deep and fine-tuning PLMs methods. Different from human intuition, the experiments demonstrated that LLMs cannot make satisfied clickbait detection just by the headlines.

Figures (1)

• Figure 1: The example of GPT-3.5 detecting clickbait errors.