Tianheng WANG

2025 Fall Ph.D.

Hi👋! I'm Tianheng WANG

Tianheng WANG is a 2025 Ph.D. student at Westlake University’s Autolab. He graduated from the University of Electronic Science and Technology of China in 2020 and obtained his master’s degree from the Department of Automation at Tsinghua University in 2023. In July 2023, he joined the Hangzhou Institute of Medicine, Chinese Academy of Sciences, as an algorithm engineer, focusing primarily on large language model training and fine-tuning. In July 2025, he joined the Autolab at Westlake University, where his research will center on large language model reasoning, vision-language model reasoning, reinforcement learning, and related areas. He has already published several papers at AI conferences such as ACL and AAAI.

Academic Achievements

Tianheng WANG has published multiple papers at several CCF-A tier international academic conferences, with primary research focuses including federated learning, natural language processing, and fine-tuning of large-scale language models. One of his representative achievements is proposing a privacy-preserving federated learning-driven framework for entity-relation extraction. This framework effectively prevents raw data leakage by introducing a task-adaptive multi-task optimization strategy at local nodes while achieving significant improvements in extraction accuracy on benchmark datasets and demonstrating better cross-domain robustness. Additionally, he innovatively applied the Kronecker product to the design of multilayer perceptron (MLP) structures. By decomposing the weight matrices of fully connected layers into low-rank tensors, this approach substantially reduces both model parameters and computational costs. In downstream tasks such as machine translation and text classification, the method maintains or even surpasses the original model’s performance while reducing parameter size by nearly an order of magnitude, offering crucial insights for the efficient design of lightweight neural network architectures.

Representative Paper

  1. Wang T, Zheng L, Lv H, et al. A distributed joint extraction framework for sedimentological entities and relations with federated learning[J]. Expert Systems with Applications, 2023, 213: 119216.
  2. Zhang J, Wang T, Wu H, et al. Sr-llm: Rethinking the structured representation in large language model[J]. arXiv preprint arXiv:2502.14352, 2025.
  3. Zhang J, Wang T, Zhang Z, et al. QiMLP: Quantum-inspired Multilayer Perceptron with Strong Correlation Mining and Parameter Compression[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2025, 39(21): 22452-22460.
  4. Zhang J, Wang T, Wang C, et al. Emotional Polarity Attention Mechanism for Text Sentiment Analysis[C]//International Conference on Database Systems for Advanced Applications. Singapore: Springer Nature Singapore, 2024: 3-18.
  5. Hou C, Liu K, Wang T, et al. DDE KG Editor: A data service system for knowledge graph construction in geoscience[J]. Geoscience Data Journal, 2024, 11(4): 1073-1085.

qzwang@163.com