Resolving Lexical Bias in Model Editing

Hammad Rizwan, Domenic Rosati, Ga Wu +1

Model editing aims to modify the outputs of large language models after they are trained. Previous approaches have often involved direct alterations to model weights, which can res…