Scholarlink

<p>Aimed at combating misinformation through Artificial Intelligence-assisted fact-checking, this paper introduces the Philippine Online Misinformation Inference (POMI) dataset, a&#xa0;compilation of 10,132 factual and false claims from Philippine fact-check articles, and the first misinformation corpus in the Philippines utilizing Natural Language Inference. In the benchmarking, LLaMA 3.1 70B achieved a&#xa0;higher accuracy of 85.1%, compared to that of smaller models such as LLaMA 3.1 8B and DeepSeek-R1 8B, as well as larger and more recent models, including LLaMA 3.3 70B and DeepSeek-R1 70B. However, applying fine-tuning to the smaller models substantially boosted their performance; in particular, LLaMA 3.1 8B increased from 78.7% to 94.5%, highlighting the competitive potential of lightweight models for NLI-based tasks. The results establish the potential of the POMI dataset to support detection of false online content in the Philippines.</p>

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

POMI: A Corpus to Support a Natural Language Inference-based Approach for Detecting Misinformation in Philippine Online News

  • 期刊论文

  • 发表日期：

    2026年2月16日

查看全文
  • Kristine Bernadette Q. Nuñez,
  • Ma. Pauline G. Abcede,
  • Led Rhoniel R. Salazar,
  • Yvonne T. Chua,
  • Victor M. Romero II,
  • Roselyn S. Gabud,
  • Paul Rossener R. Regonia

摘要

Aimed at combating misinformation through Artificial Intelligence-assisted fact-checking, this paper introduces the Philippine Online Misinformation Inference (POMI) dataset, a compilation of 10,132 factual and false claims from Philippine fact-check articles, and the first misinformation corpus in the Philippines utilizing Natural Language Inference. In the benchmarking, LLaMA 3.1 70B achieved a higher accuracy of 85.1%, compared to that of smaller models such as LLaMA 3.1 8B and DeepSeek-R1 8B, as well as larger and more recent models, including LLaMA 3.3 70B and DeepSeek-R1 70B. However, applying fine-tuning to the smaller models substantially boosted their performance; in particular, LLaMA 3.1 8B increased from 78.7% to 94.5%, highlighting the competitive potential of lightweight models for NLI-based tasks. The results establish the potential of the POMI dataset to support detection of false online content in the Philippines.