Running summarizing tools on a new article

This is the first time this article is requested and our AI summarizing tools have never been run on it. We can run our tools now if you click on the button "Run" donw the page but first make sure that it is the right article.


Knowledge Conflicts for LLMs: A Survey

Rongwu Xu, Zehan Qi, Cunxiang Wang, Hongru Wang, Yue Zhang, Wei Xu

License: CC BY-NC-ND 4.0

Abstract: This survey provides an in-depth analysis of knowledge conflicts for large language models (LLMs), highlighting the complex challenges they encounter when blending contextual and parametric knowledge. Our focus is on three categories of knowledge conflicts: context-memory, inter-context, and intra-memory conflict. These conflicts can significantly impact the trustworthiness and performance of LLMs, especially in real-world applications where noise and misinformation are common. By categorizing these conflicts, exploring the causes, examining the behaviors of LLMs under such conflicts, and reviewing available solutions, this survey aims to shed light on strategies for improving the robustness of LLMs, thereby serving as a valuable resource for advancing research in this evolving area.

Submitted to arXiv on 13 Mar. 2024

The license of this specific paper does not allow us to build upon its content and the summarizing tools will be run using the paper metadata rather than the full article. However, it still does a good job, and you can also try our tools on papers with more open licenses.