Welcome to my personal academic website

About

I am currently a Ph.D. student at Northwestern Polytechnical University, fortunate to be advised by Prof. Zhen Wang and co-advised by Dr. Qiaosheng Zhang at Shanghai AI Laboratory. Before that, I received my B.S. and M.S. degrees from Beihang University, where I was advised by Prof. Zhiguo Jiang. My research interests include graph learning, machine learning theory, multimodal large language models, and LLM safety.

News

  • πŸŽ‰ 2026: One paper accepted by ECCV 2026!
  • πŸŽ‰ 2026: One paper accepted by Neurocomputing!
  • πŸŽ‰ 2026: One paper accepted by ICML 2026!
  • πŸŽ‰ 2026: One paper accepted by ICASSP 2026!
  • πŸŽ‰ 2026: One paper accepted by Neurocomputing!
  • πŸŽ‰ 2025: One paper accepted by ICML 2025!
  • πŸŽ‰ 2024: One paper accepted by LOG 2024!

Research Interests

  • Multimodal large language models
  • Safe and trustworthy LLMs
  • LLM unlearning
  • Theory and algorithms of graph learning
  • Community detection on random graphs, e.g., stochastic block models

Education

Ph.D.
Northwestern Polytechnical University, 2022-present
M.S.
Beihang University, 2019-2022
B.S.
Beihang University, 2015-2019

Publications

Preprints

  1. Gated Graph Attention Networks with Learnable Temperature
    Z. Ma, H. Wu, Y. Zhang, Q. Zhang, Z. Wang
    arXiv preprint, 2026

Published Papers

  1. CiQi-Agent: Aligning Vision, Tools and Aesthetics in Multimodal Agent for Cultural Reasoning on Chinese Porcelains
    W. Wang*, Z. Zhou*, Z. Ma*, Y. Chen, Z. Lin, H. Sheng, P. Liu, H. Ma, W. Shao, Q. Zhang, et al.
    ECCV, 2026
    [Huggingface]
  2. Private community detection in the weighted stochastic block model
    Y. Zang, Z. Ma, Q. Zhang, Z. Wang.
    Neurocomputing, 2026
  3. MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety
    X. Wen, Z. He, H. Qi, Z. Wan, Z. Ma, Y. Wen, T. Zheng, X. Xu, C. Lu, Q. Zhang
    International Conference on Machine Learning (ICML), 2026
    [Code]
  4. Community Detection in the Multi-View Stochastic Block Model
    Y. Zhang, Z. Ma, Q. Zhang, Z. Wang, X. Li
    Neurocomputing, 2026
  5. Misclassification Rate and Privacy-Utility Trade-offs in Graph Convolutional Networks via Subsampling Stability
    Y. Zhang, Z. Ma, Q. Zhang, Z. Wang
    ICASSP, 2026
  6. Graph Attention is Not Always Beneficial: A Theoretical Analysis of Graph Attention Mechanisms via Contextual Stochastic Block Models
    Z. Ma, Q. Zhang, B. Zhou, Y. Zhang, S. Hu, Z. Wang
    International Conference on Machine Learning (ICML), 2025
    [Code]
  7. Matrix Completion with Hypergraphs: Sharp Thresholds and Efficient Algorithms
    Z. Ma, Q. Zhang, Z. Wang
    Learning on Graphs Conference (LOG), 2024
    [Code]
  8. Hyperspectral Image Classification Using Feature Fusion Hypergraph Convolution Neural Network
    Z. Ma, Z. Jiang, H. Zhang
    IEEE Transactions on Geoscience and Remote Sensing, 2022
    [Code]

* Equal contribution.

Selected Projects

Research Experience

Mar. 2025-present

Center for Safe and Trusted AI, Shanghai AI Laboratory

  • Improved LLM safety through co-evolving attacker-defender adversarial games, where attack and defense strategies are iteratively optimized against each other.
  • Worked on unlearning for LLMs with mixture-of-experts architectures, focusing on removing target knowledge or behaviors while preserving the model's general capabilities.
Oct. 2024-Apr. 2025

Collaboration with Shanghai Innovation Institute

  • Developed an agent for Chinese porcelain and antique appraisal, integrating visual understanding, tool use, and domain-specific cultural knowledge.
Apr. 2024-Oct. 2024

Frontier Research Center, Shanghai AI Laboratory

  • Conducted research on the theory of graph neural networks and attention mechanisms.

Academic Services

  • Reviewer for CVPR 2025, ECCV 2026, and ICML 2026.
Flag Counter Flag Counter