Summary of Fight Back Against Jailbreaking Via Prompt Adversarial Tuning, by Yichuan Mo et al.
Fight Back Against Jailbreaking via Prompt Adversarial Tuning
by Yichuan Mo, Yuji Wang, Zeming Wei, Yisen Wang
First submitted to arxiv on: 9 Feb 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 proposed Prompt Adversarial Tuning (PAT) method trains a prompt control attached to the user prompt as a guard prefix, optimizing it with both adversarial and benign prompts. This approach achieves intrinsic robustness in Large Language Models (LLMs), reducing the success rate of advanced attacks to nearly 0% while maintaining utility on benign tasks and incurring only negligible computational overhead. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps protect large language models from producing harmful information by using a special kind of training called Prompt Adversarial Tuning. This approach adds a special prefix to user prompts that makes it harder for attackers to trick the model into doing bad things. The result is a model that can still do its job well, but with added protection against attempts to make it say or do something harmful. |
Keywords
* Artificial intelligence * Prompt
