PRUNING AS A DEFENSE: REDUCING MEMORIZATION IN LARGE LANGUAGE MODELS

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

Track: tiny / short paper (up to 4 pages)

Keywords: pruning, memorization, LLMs

TL;DR: Our paper show that pruning reduces memorization in LLMs, making it a promising defense against membership inference attacks.

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.

Anonymization: This submission has been anonymized for double-blind review via the removal of identifying information such as names, affiliations, and identifying URLs.

Submission Number: 12