Shu Wu is currently a Full Professor and Ph.D. Supervisor at the State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation, Chinese Academy of Sciences (CASIA), China. He is also affiliated with the University of Chinese Academy of Sciences (UCAS). He received his Ph.D. degree in Computer Science from the University of Sherbrooke (UdeS), Canada, in 2012, under the supervision of Prof. Shengrui Wang. He joined CASIA in the same year.
He is a Senior Member of both IEEE and ACM, and a Distinguished Member of the China Computer Federation (CCF). He has served as an Associate Editor for ACM Transactions on Recommender Systems (ACM TORS) since 2024.
His research interests mainly include Recommendation Systems, Data Mining, and Network Data Understanding. His work has been published in top-tier conferences and journals, such as ACM TIST, IEEE TKDE, AAAI, IJCAI, SIGIR, WWW, CIKM, ICDM, WSDM, EMNLP, and ACL.
He was recognized as one of the Most Influential AI Scholars by AI2000 (2022) in the field of Information Retrieval and Recommendation, and ranked among the Top 2% of Scientists Worldwide by Stanford University. He has received several awards, including the First Prize of Beijing Natural Science Award.
Recent News
Research Overview
My research lies at the intersection of Trustworthy AI, Content Security and Graph Machine Learning. I focus on developing advanced neural architectures to model complex data structures and ensuring these systems are reliable, explainable, and aligned with human knowledge.
My academic journey has evolved across three major pillars:
1. Trustworthy LLMs & Knowledge Engineering
Currently, my lab is at the forefront of enhancing the reliability of Large Language Models (LLMs) and Multimodal Models (LVLMs).
- Knowledge Editing & Unlearning: We study how to precisely update or erase facts within LLMs without retraining, uncovering critical issues like editing-induced overfitting (ICLR '25 Spotlight) and developing robust editing frameworks (e.g., REACT, KELE).
- Hallucination Mitigation: We develop techniques to detect and steer model internal representations to reduce object hallucinations in vision-language tasks (e.g., SHARP, Logical Closed Loop).
- Alignment & Safety: Our work extends to AI safety, including machine unlearning and the detection of out-of-context or AI-generated misinformation.
2. Graph Representation Learning (GRL)
I have contributed foundational work to the GNN community, particularly in Contrastive Learning and Dynamic Graphs.
- Graph Contrastive Learning (GCL): Developed state-of-the-art frameworks for self-supervised graph learning, including adaptive augmentation strategies (WWW '21, Best Cited Paper) and structural-enhanced contrastive methods.
- Architecture Innovation: Explored disentangled representations, neighborhood interaction (GraphAIR), and graph structure learning benchmarks (GSLB) to improve the robustness and interpretability of graph models.
- Deep Graph Generation: Systematic research into the principles of generating complex, realistic graph structures.
3. User Behavior Modeling & Content Security
Applying graph and temporal modeling to solve real-world information bottlenecks and security threats.
- Graph-based Recommendation: My work on Session-based Recommendation with GNNs (AAAI '19) is a seminal paper in the field (2,200+ citations), pioneering the use of GNNs to capture transition patterns in user behavior.
- Multimedia & Social Intelligence: Researching modality-balanced learning for multimedia recommendation and temporal knowledge graph reasoning.
- Digital Forensics: Developing evidence-aware and causal intervention-based models for fake news detection and misinformation identification across social platforms.
Selected Impact & Honors
- High-Impact Publications: Author of multiple "Influential Papers" (as ranked by PaperDigest) in top-tier venues including NeurIPS, ICLR, AAAI, IJCAI, SIGIR, ACL, WWW, and KDD.
- High Citations: Several first/corresponding author works have surpassed 1,000+ citations, with a total citation count reflecting a leadership position in the Big Data and Graph Mining community.
- Broad Applications: Actively expanding AI boundaries into AI for Science (Molecular Property Prediction, 3D Molecular Representation) and FinTech (Credit Scoring).