Summary of Badllama 3: Removing Safety Finetuning From Llama 3 in Minutes, by Dmitrii Volkov
Badllama 3: removing safety finetuning from Llama 3 in minutes
by Dmitrii Volkov
First submitted to arxiv on: 1 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper investigates the subversion of Large Language Model (LLM) safety fine-tuning by an attacker with access to model weights. It evaluates three state-of-the-art fine-tuning methods – QLoRA, ReFT, and Ortho – and demonstrates how algorithmic advancements enable constant jailbreaking performance with reduced FLOPs and optimization power. The authors successfully strip safety fine-tuning from Llama 3 8B in one minute and Llama 3 70B in 30 minutes on a single GPU, suggesting ways to further reduce this timeframe. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper shows how someone can easily get around the security of big language models by accessing their weights. It looks at three methods that are commonly used – QLoRA, ReFT, and Ortho – and finds that they all have some weaknesses. By using these methods, an attacker can break into a model’s safety fine-tuning in just a few minutes. The authors also suggest ways to make this process even faster. |
Keywords
» Artificial intelligence » Fine tuning » Large language model » Llama » Optimization
