The Facade of Truth: Uncovering and Mitigating LLM Susceptibility to Deceptive Evidence
MisBelief evaluates how fabricated evidence shifts LLM belief.Abstract
This work studies how LLM internal beliefs shift under fabricated but plausible evidence. It introduces MisBelief for evaluating susceptibility to deceptive evidence and Deceptive Intent Shielding as a mitigation strategy.
Type
Publication
arXiv preprint arXiv:2601.05478
Herun Wan
CSC Visiting Student (Oct ‘25)
Visiting student; interests include Online Malicious Content Analysis such as Misinformation Detection and Social Bot Detection.
Fanxiao Li
CSC Visiting Student (Sep ‘25)
Visiting student; interests include Multimodal Misinformation, Large Vision-Language Models.
Min-Yen Kan
Associate Professor
WING lead; interests include Digital Libraries, Information Retrieval and Natural Language Processing.