We propose an efficient lifelong model editing method which maintains high performance with large number of edits.

NMKE is a fine-grained framework that uses neuron-level attribution and sparse masking for precise, efficient lifelong knowledge editing.