Pruning up to 20% of parameters in large language models (LLMs) improves their resistance to "Jailbreaking" prompts, reducing the generation of harmful and illegal content without sacrificing performance. Pruning may also enhance other LLM behaviors and improve safety and reliability.

https://arxiv.org/abs//2401.10862

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers