Summary of Assessing Adversarial Robustness Of Large Language Models: An Empirical Study, by Zeyu Yang et al.
Assessing Adversarial Robustness of Large Language Models: An Empirical Study
by Zeyu Yang, Zhao Meng, Xiaochen Zheng, Roger Wattenhofer
First submitted to arxiv on: 4 May 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
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 abstract presents a novel white-box attack approach that targets vulnerabilities in leading open-source Large Language Models (LLMs). The method assesses the impact of model size, structure, and fine-tuning strategies on the resistance of LLMs like Llama, OPT, and T5 to adversarial perturbations. The study evaluates the robustness of these models across five text classification tasks, establishing a new benchmark for LLM reliability. This research has significant implications for the trustworthy deployment of LLMs in real-world applications. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large Language Models (LLMs) are super smart computers that can understand and generate human-like language. But hackers have found ways to trick them into saying things they don’t mean. This paper shows how we can create attacks that fool these models, even when they’re really good. We tested different sizes and types of LLMs and found out what makes them more or less resistant to these tricks. The results will help us make sure these powerful tools are reliable and safe for real-life use. |
Keywords
* Artificial intelligence * Fine tuning * Llama * T5 * Text classification
