Dissecting Large Language Models

Published: 23 Oct 2023, Last Modified: 28 Nov 2023SoLaR PosterEveryoneRevisionsBibTeX
Keywords: pruning, machine unlearning, capability removal, intepretability, modularity
TL;DR: We propose a method for machine unlearning via pruning, and find that language models have task specialized neurons.
Abstract: Understanding and shaping the behaviour of Large Language Models (LLMs) is increasingly important as applications become more powerful and more frequently adopted. This paper introduces a machine unlearning method specifically designed for LLMs. We introduce a selective pruning method for LLMs that removes neurons based on their relative importance on a targeted capability compared to overall network performance. This approach is a compute- and data-efficient method for identifying and removing neurons that enable specific behaviours. Our findings reveal that both feed-forward and attention neurons in LLMs are specialized; that is, for specific tasks, certain neurons are more crucial than others.
Submission Number: 26