A Survey on Knowledge Conflicts in the Era of LLMs

Anonymous

Abstract: This survey presents a comprehensive examination of knowledge conflicts in Large Language Models (LLMs). It explores the intricate challenges that arise when LLMs integrate contextual knowledge with their parametric knowledge. Our focus is on three primary types of knowledge conflicts: context-memory, inter-context, and intra-memory conflict. These conflicts can significantly impact the trustworthiness and accuracy of LLMs, especially in real-world applications where misinformation and noise are prevalent. The survey categorizes these conflicts, investigates their causes, and reviews potential mitigation strategies. It aims to provide insights into enhancing the robustness of LLMs, making it a valuable resource for advancing research in this evolving area.

Paper Type: long

Research Area: Interpretability and Analysis of Models for NLP

Contribution Types: Surveys

Languages Studied: English