Zihao Zhou

Zihao Zhou 周梓浩
Google Scholar / Github / Twitter
zihao.zhou@liverpool.ac.uk

Hi! I am a CS PhD student at University of Liverpool and Xi’an Jiaotong-liverpool University, beginning from 2022 Fall. My supervisory team includes Prof. Qiufeng Wang, Prof. Kaizhu Huang and Prof. Xiaowei Huang.

My research interests lie in:

• Large Language Models Reasoning: Evaluating and improving the reasoning ability of LLMs, particularly in math solving ability.
• Mathematical Question Answering : Math word problem, Geometry problem, etc.
• Dialog tutoring System: Exploring how to build tutoring system like real teacher.

Publications

• [ICLR 2025] Is Your Model Really A Good Math Reasoner? Evaluating Mathematical Reasoning with Checklist. [website] [paper] [code]
  • Zihao Zhou*, Shudong Liu*, Maizhen Ning, Wei Liu, Jingdong Wang, Derek F. Wong, Xiaowei Huang, Qiufeng Wang, Kaizhu Huang
  • 🔍 MathCheck includes multiple mathematical reasoning tasks and robustness test types to facilitate a comprehensive evaluation of both mathematical reasoning ability and behavior testing.
    
    
• [AAAI 2025] GNS: Solving Plane Geometry Problems by Neural-Symbolic Reasoning with Multi-Modal LLMs.
  • Maizhen Ning*, Zihao Zhou*, Qiufeng Wang, Xiaowei Huang, Kaizhu Huang
  • GNS is a neural-symbolic reasoning system simultaneously capable of multiple geometry reasoning tasks beyond solving. Training: multi-tasks SFT. Inference: combining knowledge prediction, geometry image parsing and tool calling.
  • Selected as Oral Presentation(Top 5%) 🎉.
    
    
• [AAAI 2024] MathAttack: Attacking Large Language Models Towards Math Solving Ability. [paper] [code]
  • Zihao Zhou, Qiufeng Wang, Mingyu Jin, Jie Yao, Jianan Ye, Wei Liu, Wei Wang, Xiaowei Huang, Kaizhu Huang
  • Examining the robustness of LLMs in math reasoning ability by textual attack algorithm.
    
    
• [ACL 2023 Findings] Learning by Analogy: Diverse Questions Generation in Math Word Problem. [paper] [code]
  • Zihao Zhou*, Maizhen Ning*, Qiufeng Wang, Jie Yao, Wei Wang, Xiaowei Huang and Kaizhu Huang
  • ✍️ A framework to generate diverse questions with labels for a given math word problems.
    
    
• [NLPCC 2023] Solving Math Word Problem with Problem Type Classification. [paper][code]
  • Jie Yao*, Zihao Zhou*, Qiufeng Wang
  • A problem type classifier to combine the abilities of LLM solver and traditional math solver.
  • 🎯 Ranked 2nd in NLPCC 2023 Share Task 3.
    
    
• [ICASSP 2021] Knowledge-Based Chat Detection with False Mention Discrimination. [paper]
  • Wei Liu, Peijie Huang, Dongzhu Liang, Zihao Zhou
  • A two-stage pipeline for chat detection 🗣️🤖.

Experience

• Tsinghua University, Visiting Student at KEG Lab, 2023.09 - 2024.03
• University of Liverpool, PhD Student, 2022.09 - Present
• South China Agricultural University, Bachelor Student, 2018.09 - 2022.06

Services

• Conference Reviewer: NeurIPS 2024, ICML 2024, ACL ARR 2023/2024
• Journal Reviewer: Neural Networks