Cutting Off the Head Ends the Conflict: A Mechanism for Interpreting and Mitigating Knowledge Conflicts in Language Models

arXiv:2402.18154v1 Announce Type: new
Abstract: Recently, retrieval augmentation and tool augmentation have demonstrated a remarkable capability to expand the internal memory boundaries of language models (LMs) by providing external context. However, internal memory and external context inevitably clash, leading to knowledge conflicts within LMs. In this paper, we aim to interpret the mechanism of knowledge conflicts through the lens of information flow, and then mitigate conflicts by precise interventions at the pivotal point. We find there are some attention heads …

abstract arxiv augmentation capability conflict context cs.ai cs.cl cs.ir expand head knowledge language language models lms memory retrieval tool type