Summary of Efficient Llm Jailbreak Via Adaptive Dense-to-sparse Constrained Optimization, by Kai Hu et al.
Efficient LLM Jailbreak via Adaptive Dense-to-sparse Constrained Optimization
by Kai Hu, Weichen Yu, Yining Li, Kai Chen, Tianjun Yao, Xiang Li, Wenhe Liu, Lijun Yu, Zhiqiang Shen, Matt Fredrikson
First submitted to arxiv on: 15 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 A novel method, Adaptive Dense-to-Sparse Constrained Optimization (ADC), is introduced to successfully jailbreak multiple open-source large language models (LLMs). This approach relaxes discrete token optimization into a continuous process, gradually increasing the sparsity of optimizing vectors. Experimental results show that ADC outperforms state-of-the-art token-level methods, achieving the highest attack success rate on seven out of eight LLMs on Harmbench. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models can create harmful content when attacked. Researchers have developed a new way to make these models do bad things. They call it “jailbreaking.” This paper shows how they did it using a method called Adaptive Dense-to-Sparse Constrained Optimization (ADC). It’s like finding a shortcut between two points, making the process more efficient. |
Keywords
» Artificial intelligence » Optimization » Token
