MathAttack: Attacking Large Language Models towards Math Solving Ability

doi:10.1609/aaai.v38i17.29949

	MathAttack: Attacking Large Language Models towards Math Solving Ability
	Zhou, Zihao 1,2; Wang, Qiufeng 1; Jin, Mingyu 3; Yao, Jie 1,2; Ye, Jianan 1,2; Liu, Wei4 ; Wang, Wei 1; Huang, Xiaowei 2; Huang, Kaizhu 5
	2024-03-25
会议录名称	PROCEEDINGS OF THE AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE
ISSN	2159-5399
卷号	38
期号	17
页码	19750-19758
DOI	10.1609/aaai.v38i17.29949
摘要	With the boom of Large Language Models (LLMs), the research of solving Math Word Problem (MWP) has recently made great progress. However, there are few studies to examine the robustness of LLMs in math solving ability. Instead of attacking prompts in the use of LLMs, we propose a MathAttack model to attack MWP samples which are closer to the essence of robustness in solving math problems. Compared to traditional text adversarial attack, it is essential to preserve the mathematical logic of original MWPs during the attacking. To this end, we propose logical entity recognition to identify logical entries which are then frozen. Subsequently, the remaining text are attacked by adopting a word-level attacker. Furthermore, we propose a new dataset RobustMath to evaluate the robustness of LLMs in math solving ability. Extensive experiments on our RobustMath and two another math benchmark datasets GSM8K and MultiAirth show that MathAttack could effectively attack the math solving ability of LLMs. In the experiments, we observe that (1) Our adversarial samples from higher-accuracy LLMs are also effective for attacking LLMs with lower accuracy (e.g., transfer from larger to smaller-size LLMs, or from few-shot to zero-shot prompts); (2) Complex MWPs (such as more solving steps, longer text, more numbers) are more vulnerable to attack; (3) We can improve the robustness of LLMs by using our adversarial samples in few-shot prompts. Finally, we hope our practice and observation can serve as an important attempt towards enhancing the robustness of LLMs in math solving ability. The code and dataset is available at: https://github.com/zhouzihao501/MathAttack. © 2024, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
会议录编者/会议主办者	Association for the Advancement of Artificial Intelligence
关键词	Computational linguistics Benchmark datasets Entity recognition High-accuracy Language model Word level Word problem
会议名称	38th AAAI Conference on Artificial Intelligence, AAAI 2024
会议地点	Vancouver, BC, Canada
会议日期	February 20, 2024 - February 27, 2024
URL	查看原文
收录类别	EI
语种	英语
出版者	Association for the Advancement of Artificial Intelligence
EI入藏号	20241515874316
EI主题词	Zero-shot learning
EISSN	2374-3468
EI分类号	721.1 Computer Theory, Includes Formal Logic, Automata Theory, Switching Theory, Programming Theory
原始文献类型	Conference article (CA)
文献类型	会议论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/364711
专题	信息科学与技术学院_硕士生
通讯作者	Wang, Qiufeng
作者单位	1.School of Advanced Technology, Xi’an Jiaotong-Liverpool University, China 2.University of Liverpool, United Kingdom 3.Northwestern University, United States 4.ShanghaiTech University, China 5.Duke Kunshan University, China
推荐引用方式 GB/T 7714	Zhou, Zihao,Wang, Qiufeng,Jin, Mingyu,et al. MathAttack: Attacking Large Language Models towards Math Solving Ability[C]//Association for the Advancement of Artificial Intelligence:Association for the Advancement of Artificial Intelligence,2024:19750-19758.