Detecting Edited Knowledge in Language Models

arXiv:2405.02765v1 Announce Type: new
Abstract: Knowledge editing techniques (KEs) can update language models' obsolete or inaccurate knowledge learned from pre-training. However, KE also faces potential malicious applications, e.g. inserting misinformation and toxic content. Moreover, in the context of responsible AI, it is instructive for end-users to know whether a generated output is driven by edited knowledge or first-hand knowledge from pre-training. To this end, we study detecting edited knowledge in language models by introducing a novel task: given an edited …

abstract applications arxiv context cs.ai cs.cl editing generated however knowledge language language models misinformation pre-training responsible responsible ai training type update