About Me

I am a postdoctoral researcher at Great Bay University and the University of Science and Technology of China, supervised by Prof. Zaiwen Wen and Prof. Dong Liu, and a guest researcher at Leiden University. I received my Ph.D. in Computer Science from Leiden in 2025, supervised by Prof. Matthijs van Leeuwen and Prof. Thomas Bäck. During my Ph.D., I visited the DAML group at TU Munich, hosted by Prof. Stephan Günnemann. I studied statistics at Tongji University and data science at ENSAI, France.

My research focuses on two areas: AI for operations research (AI4OR) and trustworthy data mining. In AI4OR, I develop methods based on large language and multimodal models (LLMs/MLLMs) for automated optimization modeling: converting natural-language and visual problem descriptions into mathematical models and executable solver code. I develop benchmarks to evaluate modeling systems and investigate post-training, including supervised fine-tuning and reinforcement learning, to improve them. I am also interested in their trustworthiness—robustness, explainability, privacy, and hallucination detection and mitigation—to support reliable use. My broader interests extend to LLM/MLLM-based optimization solving, including algorithm design and solver configuration, and its interaction with modeling. As my research has expanded toward AI4OR, I continue my Ph.D. line of work in trustworthy data mining, developing interpretable, label-efficient, robust, and transferable methods for anomaly detection, graph learning, and subgroup discovery, as well as diffusion and flow matching for tabular data, with applications in smart manufacturing and health management. My doctoral research, Trustworthy Anomaly Detection for Smart Manufacturing, was part of the NWO Digital Twin programme and focused on predictive maintenance for lithography wafer-handling systems and high-end printers, in collaboration with ASML, VDL-ETG, and Canon. Before my Ph.D., I worked on lithography systems at Shanghai Micro Electronics Equipment and financial compliance analytics at HSBC France.

My work has appeared at NeurIPS, ICML, AAAI, TKDE, TKDD, and DMKD, including a NeurIPS Spotlight, an AAAI oral presentation, and an ESI Highly Cited Paper. It has received 635 Google Scholar citations (as of 13 October 2026). I was selected for the Chinese Ministry of Education’s overseas postdoctoral talent programme in 2025 (approximately CNY 1 million in funding for 2025–2028) and received the Eiffel Excellence Scholarship (approximately CNY 250,000). I also serve as an Area Chair for the AI for Math Workshop at ICML 2026 and a Session Co-Chair at AAAI 2026.

Research Interests

Latest news

Does Compression Imply Generalization? A Minimum Description Length Perspective is accepted at NeurIPS 2026. Congratulations to Lincen!

Optimal Subgroup Discovery at Every Support Threshold is accepted at NeurIPS 2026 as a Spotlight. Congratulations to Lincen and Qi!

Efficient Flow Matching for Sparse-View CT Reconstruction is accepted at MICCAI 2026. Congratulations to Jiayang!

Constructing Industrial-Scale Optimization Modeling Benchmark is accepted at ICML 2026.

Earlier updates (2021–2025)

Towards Automated Self-Supervised Learning for Truly Unsupervised Graph Anomaly Detection is selected for presentation in the AAAI 2026 Journal Track.

Learning Subgroups with Maximum Treatment Effects without Causal Heuristics is accepted at AAAI 2026 for an oral presentation. Congratulations to Lincen!

Successfully defended my Ph.D. thesis, Trustworthy Anomaly Detection for Smart Manufacturing, at Leiden University!

Presented Explainable Graph Neural Networks Under Fire at the Amsterdam meetup for the global LoG Conference.

Cross-Domain Graph Level Anomaly Detection is accepted by TKDE.

Graph Neural Networks based Log Anomaly Detection and Explanation is accepted for the ICSE 2024 Poster Track.

Position cancelled. Internship and thesis opportunity with Canon on Inkjet Jet Failure Detection.

A Survey on Explainable Anomaly Detection is accepted by TKDD. Recognized as an ESI Highly Cited Paper.

Position filled. Internship and thesis opportunity with VDL & ASML on Predictive Maintenance via Log Anomaly Detection.

Publications

# Equal contribution* Corresponding author

Google Scholar ↗
Loading publications…

Education & Experience

Academic Positions

Postdoctoral Researcher

Great Bay University & USTC

Supervisors: Zaiwen Wen · Dong Liu

Guest Researcher

Leiden University

Visiting Ph.D. Researcher

Technical University of Munich

Host: Stephan Günnemann

Education

Diplôme d’Ingénieur in Data Science

ENSAI, France

Eiffel Excellence Scholarship

M.Sc. in Statistics

Tongji University

Supervisor: Hanying Liang

B.Sc. in Statistics

Tongji University

Supervisor: Hanying LiangShanghai Outstanding Graduate

Academic service

Chairing

Area Chair, AI for Math Workshop @ ICML 2026
Session Co-Chair, AAAI 2026 · Data Mining & Knowledge Management

Reviewing

Conferences: NeurIPS (2026) · ICML (2026) · ICLR (2024–2025) · AAAI (2024–2025) · CVPR (2026) · ECCV (2026) · KDD (2024, 2026) · ECML-PKDD (2025–2026)

Journals: TKDE · DMKD · IJCV · TDSC · TETCI · TCC · IEEE Internet of Things Journal (IoT-J) · IEEE Intelligent Systems · Pattern Recognition · Computer Networks · AI Communications

Mentoring

Master’s students

  • Yuhang Wang

    Unsupervised Model Selection for Graph Representation Learning

    – · Leiden University · Cum Laude

  • Timo Vos

    Predictive Maintenance of Lithography System Using Log Anomaly Detection

    – · Leiden University
    In collaboration with VDL-ETG and ASML

  • Yuxuan Zhu

    Data-driven inkjet jet failure detection and classification using piezo self-sensing

    – · Leiden University
    In collaboration with Canon Production Printing

Bachelor’s students

  • Samuel Smulders

    Automated feature selection for unsupervised anomaly detection with internal evaluation strategies

    – · Leiden University

  • Milan Luijken

    The effects of feature selection on anomaly detection

    – · Leiden University

Contact

WeChat: DigitalTwinNL

Email: z.li(at)liacs.leidenuniv.nl

Cite this work