Summary of Evaluating Robustness Of Generative Search Engine on Adversarial Factual Questions, by Xuming Hu et al.
Evaluating Robustness of Generative Search Engine on Adversarial Factual Questions
by Xuming Hu, Xiaochuan Li, Junzhe Chen, Yinghui Li, Yangning Li, Xiaoguang Li, Yasheng Wang, Qun Liu, Lijie Wen, Philip S. Yu, Zhijiang Guo
First submitted to arxiv on: 25 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
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 This research paper proposes evaluating the robustness of generative search engines in a high-risk setting, where adversaries aim to deceive the model into returning incorrect responses. The study focuses on Bing Chat, PerplexityAI, and YouChat, examining their performance across diverse queries using adversarial factual questions. The results show that retrieval-augmented generation is more susceptible to factual errors compared to large language models (LLMs) without retrieval. This highlights potential security risks of these systems, emphasizing the need for rigorous evaluation before deployment. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Generative search engines have the power to change how people find information online. But current models may not always give accurate answers. To make things worse, they can be tricked into giving wrong answers by someone trying to deceive them. This research checks how well generative search engines do when faced with realistic and high-risk situations where someone is trying to cheat the system. The study shows that some systems are more likely to get questions wrong than others. This matters because it could lead to people getting bad information. |
Keywords
» Artificial intelligence » Retrieval augmented generation
