Towards unified multimodal editing with enhanced knowledge collaboration

• Main Authors: PAN, Kaihang, FAN, Zhaoyu, LI, Juncheng, YU, Qifan, FEI, Hao, TANG, Siliang, HONG, Richang, ZHANG, Hanwang, Qianru SUN
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2024
• Subjects: Multimodal LLMs; Knowledge editing; Intrinsic knowledge editing; External knowledge resorting; Databases and Information Systems
• Online Access: https://ink.library.smu.edu.sg/sis_research/9401; https://ink.library.smu.edu.sg/context/sis_research/article/10401/viewcontent/2409.19872v2__2_.pdf
• Institution: Singapore Management University

The swift advancement in Multimodal LLMs (MLLMs) also presents significant challenges for effective knowledge editing. Current methods, including intrinsic knowledge editing and external knowledge resorting, each possess strengths and weaknesses, struggling to balance the desired properties of reliability, generality, and locality when applied to MLLMs. In this paper, we propose UniKE, a novel multimodal editing method that establishes a unified perspective and paradigm for intrinsic knowledge editing and external knowledge resorting. Both types of knowledge are conceptualized as vectorized key-value memories, with the corresponding editing processes resembling the assimilation and accommodation phases of human cognition, conducted at the same semantic levels. Within such a unified framework, we further promote knowledge collaboration by disentangling the knowledge representations into the semantic and truthfulness spaces.