| Aug 29, 2025 |   We propose MMSearch-Plus — a challenging MLLM agent benchmark that requires high-level multimodal understanding and tool-use capability through iterative image-text retrieval and cross-validation under retrieval noise. Released benchmark dataset is available here.  |  
  | Jul 14, 2025 |   We propose Guided Hybrid Policy Optimization (GHPO) — a difficulty-aware RLVR framework that improves training stability and reasoning performance in LLMs, achieving strong gains on challenging math benchmarks. Code is available here.  |  
  | Oct 28, 2024 |   I have joined Huawei HKRC in Hong Kong as a full-time research engineer.  |  
  | May 21, 2024 |   Happy to announce that I have passed my PhD oral defense!  |  
  | May 17, 2024 |   One paper accepted by KDD 2024: Geometric View of Soft Decorrelation in Self-Supervised Learning. Congratulations to Yifei!  |  
  | Apr 24, 2024 |   Another survey paper accepted by ACM TIST: A Survey of Trustworthy Federated Learning: Issues, Solutions, and Challenges. Congratulations to Yifei!  |  
  | Apr 20, 2024 |   One survey paper accepted by IJCAI 2024: A Systematic Survey on Federated Semi-supervised Learning. Congratulations to Zixing!  |  
  | Nov 14, 2023 |   One paper accepted by Neural Networks: MECCH: Metapath Context Convolution-based Heterogeneous Graph Neural Networks  |  
  | Jun 18, 2023 |   Our tutorial Trustworthy Federated Learning: Concepts, Methods, Applications, and Beyond was presented at IJCNN 2023  |  
  | Apr 20, 2023 |   One paper accepted by IJCAI 2023: FedHGN: A Federated Framework for Heterogeneous Graph Neural Networks  |  
  | Apr 04, 2023 |   Served as a reviewer in the WWW 2023 “Artifacts Available” review committee  |  
  | Jan 14, 2023 |   Fellow students from our MISC Lab presented our recent research on LOGS  |  
  | Jul 28, 2022 |   Our tutorial Deep Learning on Graphs was presented at DeepLearn 2022 Summer  |  
  | Nov 22, 2020 |   Our tutorial Deep Learning on Graphs: Methods and Applications was presented at ICONIP 2020  |  
  | Jan 11, 2020 |   One paper accepted by WWW 2020: MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding  |