EUVIMEDEUVIMED entdecken
Crossref

CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG

Boyi Deng, Wenjie Wang, Fengbin Zhu, Qifan Wang, Fuli Feng

Proceedings of the AAAI Conference on Artificial Intelligence · 11.04.2025 · 39 (22) · 23760-23768

DOI: 10.1609/aaai.v39i22.34547 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 11.04.2025. Verfasst von Boyi Deng und 4 weiteren Autor:innen. Die Arbeit wurde bisher 11-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Retrieval-Augmented Generation (RAG) can alleviate hallucinations of Large Language Models (LLMs) by referencing external documents. However, the misinformation in external documents may mislead LLMs' generation. To address this issue, we explore the task of "credibility-aware RAG", in which LLMs automatically adjust the influence of retrieved documents based on their credibility scores to counteract misinformation. To this end, we introduce a plug-and-play method named Credibility-aware Attention Modification (CrAM). CrAM identifies influential attention heads in LLMs and adjusts their attention weights based on the credibility of the documents, thereby reducing the impact of low-credibility documents. Experiments on Natual Questions and TriviaQA using Llama2-13B, Llama3-8B, and Qwen1.5-7B show that CrAM improves the RAG performance of LLMs against misinformation pollution by over 20%, even surpassing supervised fine-tuning methods.

Zitieren

APA

Deng, B., Wang, W., Zhu, F., Wang, Q., & Feng, F. (2025). CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG. Proceedings of the AAAI Conference on Artificial Intelligence, 39(22), 23760-23768. https://doi.org/10.1609/aaai.v39i22.34547

Vancouver

Deng B, Wang W, Zhu F, Wang Q, Feng F. CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG. Proceedings of the AAAI Conference on Artificial Intelligence. 2025;39(22):23760-23768. doi:10.1609/aaai.v39i22.34547

BibTeX
@article{Deng2025,
  author = {Deng, Boyi and Wang, Wenjie and Zhu, Fengbin and Wang, Qifan and Feng, Fuli},
  title = {CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2025},
  volume = {39},
  number = {22},
  pages = {23760--23768},
  doi = {10.1609/aaai.v39i22.34547}
}
Kostenlos starten →