Evaluating Robustness of Generative Search Engine on Adversarial Factual Questions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hu, Xuming, Li, Xiaochuan, Chen, Junzhe, Li, Yinghui, Li, Yangning, Li, Xiaoguang, Wang, Yasheng, Liu, Qun, Wen, Lijie, Yu, Philip S., Guo, Zhijiang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914718964252672
author Hu, Xuming
Li, Xiaochuan
Chen, Junzhe
Li, Yinghui
Li, Yangning
Li, Xiaoguang
Wang, Yasheng
Liu, Qun
Wen, Lijie
Yu, Philip S.
Guo, Zhijiang
author_facet Hu, Xuming
Li, Xiaochuan
Chen, Junzhe
Li, Yinghui
Li, Yangning
Li, Xiaoguang
Wang, Yasheng
Liu, Qun
Wen, Lijie
Yu, Philip S.
Guo, Zhijiang
contents Generative search engines have the potential to transform how people seek information online, but generated responses from existing large language models (LLMs)-backed generative search engines may not always be accurate. Nonetheless, retrieval-augmented generation exacerbates safety concerns, since adversaries may successfully evade the entire system by subtly manipulating the most vulnerable part of a claim. To this end, we propose evaluating the robustness of generative search engines in the realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning incorrect responses. Through a comprehensive human evaluation of various generative search engines, such as Bing Chat, PerplexityAI, and YouChat across diverse queries, we demonstrate the effectiveness of adversarial factual questions in inducing incorrect responses. Moreover, retrieval-augmented generation exhibits a higher susceptibility to factual errors compared to LLMs without retrieval. These findings highlight the potential security risks of these systems and emphasize the need for rigorous evaluation before deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2403_12077
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Evaluating Robustness of Generative Search Engine on Adversarial Factual Questions
Hu, Xuming
Li, Xiaochuan
Chen, Junzhe
Li, Yinghui
Li, Yangning
Li, Xiaoguang
Wang, Yasheng
Liu, Qun
Wen, Lijie
Yu, Philip S.
Guo, Zhijiang
Computation and Language
Artificial Intelligence
Information Retrieval
Generative search engines have the potential to transform how people seek information online, but generated responses from existing large language models (LLMs)-backed generative search engines may not always be accurate. Nonetheless, retrieval-augmented generation exacerbates safety concerns, since adversaries may successfully evade the entire system by subtly manipulating the most vulnerable part of a claim. To this end, we propose evaluating the robustness of generative search engines in the realistic and high-risk setting, where adversaries have only black-box system access and seek to deceive the model into returning incorrect responses. Through a comprehensive human evaluation of various generative search engines, such as Bing Chat, PerplexityAI, and YouChat across diverse queries, we demonstrate the effectiveness of adversarial factual questions in inducing incorrect responses. Moreover, retrieval-augmented generation exhibits a higher susceptibility to factual errors compared to LLMs without retrieval. These findings highlight the potential security risks of these systems and emphasize the need for rigorous evaluation before deployment.
title Evaluating Robustness of Generative Search Engine on Adversarial Factual Questions
topic Computation and Language
Artificial Intelligence
Information Retrieval
url https://arxiv.org/abs/2403.12077