MathAttack: Attacking Large Language Models towards Math Solving Ability

	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
会议录名称	THIRTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 38 NO 17
ISSN	2159-5399
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.
会议名称	38th AAAI Conference on Artificial Intelligence (AAAI) / 36th Conference on Innovative Applications of Artificial Intelligence / 14th Symposium on Educational Advances in Artificial Intelligence
出版地	2275 E BAYSHORE RD, STE 160, PALO ALTO, CA 94303 USA
会议地点	null,Vancouver,CANADA
会议日期	FEB 20-27, 2024
URL	查看原文
收录类别	CPCI-S
语种	英语
资助项目	National Natural Science Foundation of China["92370119","62376113","62276258"] ; Jiangsu Science and Technology Programme (Natural Science Foundation of Jiangsu Province)[BE2020006-4] ; European Union[956123] ; UK EPSRC[EP/T026995/1]
WOS研究方向	Computer Science ; Education & Educational Research
WOS类目	Computer Science, Artificial Intelligence ; Computer Science, Interdisciplinary Applications ; Computer Science, Theory & Methods ; Education, Scientific Disciplines
WOS记录号	WOS:001239407300139
出版者	ASSOC ADVANCEMENT ARTIFICIAL INTELLIGENCE
EISSN	2374-3468
引用统计	正在获取...
文献类型	会议论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/381359
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.Xian Jiaotong Liverpool Univ, Sch Adv Technol, Suzhou, Jiangsu, Peoples R China 2.Univ Liverpool, Liverpool, Merseyside, England 3.Northwestern Univ, Evanston, IL 60208 USA 4.ShanghaiTech Univ, Shanghai, Peoples R China 5.Duke Kunshan Univ, Suzhou, Jiangsu, Peoples R 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]. 2275 E BAYSHORE RD, STE 160, PALO ALTO, CA 94303 USA:ASSOC ADVANCEMENT ARTIFICIAL INTELLIGENCE,2024.