Knowledge Conflict
Knowledge conflict in large language models (LLMs) arises from discrepancies between the model's internal knowledge and information provided in the context, or between multiple conflicting sources. Current research focuses on detecting and resolving these conflicts, employing techniques like contrastive decoding, adaptive decoding methods, and attention mechanism adjustments within various model architectures including LLMs and vision-language models. Understanding and mitigating knowledge conflicts is crucial for improving the reliability and trustworthiness of LLMs, particularly in applications requiring factual accuracy and robust reasoning under uncertainty.
Papers
Analysing the Residual Stream of Language Models Under Knowledge Conflicts
Yu Zhao, Xiaotang Du, Giwon Hong, Aryo Pradipta Gema, Alessio Devoto, Hongru Wang, Xuanli He, Kam-Fai Wong, Pasquale Minervini
Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
Yu Zhao, Alessio Devoto, Giwon Hong, Xiaotang Du, Aryo Pradipta Gema, Hongru Wang, Kam-Fai Wong, Pasquale Minervini
Adaptive Question Answering: Enhancing Language Model Proficiency for Addressing Knowledge Conflicts with Source Citations
Sagi Shaier, Ari Kobren, Philip Ogren
ECon: On the Detection and Resolution of Evidence Conflicts
Cheng Jiayang, Chunkit Chan, Qianqian Zhuang, Lin Qiu, Tianhang Zhang, Tengxiao Liu, Yangqiu Song, Yue Zhang, Pengfei Liu, Zheng Zhang