Scholarlink

Hallucinations are a known barrier in using generative AI in education. This paper proposes to use the troublemaker agent strategy in order to turn these flaws into productive assets. We present the Troublemaker Agent Platform (TrAP), which supports this approach and includes a low-cost, eco-friendly architecture based on RAG and SLM for independence from commercial LLMs.

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

Making Generative AI Hallucinations Useful by Reassessing the Troublemaker Agent Strategy

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Emmanuel G. Blanchard,
  • Jean-Christophe Callahan,
  • Idir Akretche,
  • Naif Asswiel,
  • Lucas Schmitt,
  • Amine Kessemtini,
  • Imran S. A. Khan

摘要

Hallucinations are a known barrier in using generative AI in education. This paper proposes to use the troublemaker agent strategy in order to turn these flaws into productive assets. We present the Troublemaker Agent Platform (TrAP), which supports this approach and includes a low-cost, eco-friendly architecture based on RAG and SLM for independence from commercial LLMs.