all InfoSec News
Badllama 3: removing safety finetuning from Llama 3 in minutes
July 2, 2024, 4:14 a.m. | Dmitrii Volkov
cs.CR updates on arXiv.org arxiv.org
Abstract: We show that extensive LLM safety fine-tuning is easily subverted when an attacker has access to model weights. We evaluate three state-of-the-art fine-tuning methods-QLoRA, ReFT, and Ortho-and show how algorithmic advances enable constant jailbreaking performance with cuts in FLOPs and optimisation power. We strip safety fine-tuning from Llama 3 8B in one minute and Llama 3 70B in 30 minutes on a single GPU, and sketch ways to reduce this further.
access art arxiv attacker cs.ai cs.cl cs.cr cs.lg enable fine-tuning finetuning jailbreaking llama llm performance power safety state
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
Software Engineer
@ Booz Allen Hamilton | USA, VA, McLean (8283 Greensboro Dr, Hamilton)
SOC Level 1 Engineer
@ Groupon | Remote - India
Senior Technology Auditor (Continuous Process Monitoring)
@ CNA Insurance | US- IL40- Chicago-151N Frankln
Sr. Director, Tech Process Management (ES Risk)
@ Capital One | McLean, VA
AVP, Pre-Sales and Professional Services for Group Benefits & Affinity
@ Manulife | CAN, Ontario, Toronto, 250 Bloor Street East
Software Engineer III
@ Walmart | IN KA BANGALORE Home Office PW II