Poster
in
Workshop: Building Trust in LLMs and LLM Applications: From Guardrails to Explainability to Regulation

PRUNING AS A DEFENSE: REDUCING MEMORIZATION IN LARGE LANGUAGE MODELS

Mansi Gupta · Nikhar Waghela · Sarthak Gupta · Shourya Goel · Sanjif Shanmugavelu

Project Page [ OpenReview]

Abstract

Large language models have been shown to memorize significant portions of their training data, which they can reproduce when appropriately prompted. This work investigates the impact of simple pruning techniques on this behavior. Our findings reveal that pruning effectively reduces the extent of memorization in LLMs, demonstrating its potential as a foundational approach for mitigating membership inference attacks.

Chat is not available.