Scholarlink

Detecting clickbait – those headlines that attract excessive attention – remains an interesting topic in natural language processing. This paper presents an applied approach to categorizing clickbait texts by fine-tuning large-scale language models, that is, Mistral 7B and LLaMA 3.1 8B, with an algorithm of PEFT coupled with QLoRA. The approach is evaluated on the Webis Clickbait Corpus 2017 dataset. To get an overall evaluation, we also contrast our model with multiple prompting methods. The experiment demonstrates that the fine-tuning techniques not only increase the correct rate but also provide better insight into the reason why the model makes the judgment.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fine-Tuning Large Language Models for Clickbait Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nguyen Hong Vu,
  • Nguyen Phuoc Dai

摘要

Detecting clickbait – those headlines that attract excessive attention – remains an interesting topic in natural language processing. This paper presents an applied approach to categorizing clickbait texts by fine-tuning large-scale language models, that is, Mistral 7B and LLaMA 3.1 8B, with an algorithm of PEFT coupled with QLoRA. The approach is evaluated on the Webis Clickbait Corpus 2017 dataset. To get an overall evaluation, we also contrast our model with multiple prompting methods. The experiment demonstrates that the fine-tuning techniques not only increase the correct rate but also provide better insight into the reason why the model makes the judgment.