Summary of Stepwise Reasoning Error Disruption Attack Of Llms, by Jingyu Peng et al.
Stepwise Reasoning Error Disruption Attack of LLMs
by Jingyu Peng, Maolin Wang, Xiangyu Zhao, Kai Zhang, Wanyu Wang, Pengyue Jia, Qidong Liu, Ruocheng Guo, Qi Liu
First submitted to arxiv on: 16 Dec 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- 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 attack on large language models (LLMs) is proposed, which subtly injects errors into prior reasoning steps to mislead the model and produce incorrect answers. The Stepwise rEasoning Error Disruption (SEED) attack is compatible with zero-shot and few-shot settings, maintains the natural reasoning flow, and ensures covert execution without modifying the instruction. Extensive experiments on four datasets across four different models demonstrate SEED’s effectiveness in revealing the vulnerabilities of LLMs to disruptions in reasoning processes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models have made big progress in solving complex problems, but their ability to make safe and good decisions is still not fully understood. A new way to attack these models is discovered, which makes them produce wrong answers by introducing small mistakes into their thought process. This attack works even when the model doesn’t see the correct answer before, and it does so without changing what the question is asking. |
Keywords
» Artificial intelligence » Few shot » Zero shot
