#model editing

EPFL researchers have developed MEMOIR, a novel framework that enables continuous, reliable, and localized updates in large language models, outperforming existing methods in various benchmarks.