Liangyu Wu (吴亮宇)

Undergraduate Researcher in Artificial Intelligence
Peking University, Yuanpei College

📍 Beijing, China
✉️ zmwandering@gmail.com

profile

👋 About Me

I am a senior undergraduate at Peking University (Yuanpei College) majoring in Artificial Intelligence, a member of the Tong Class program.

My research lies at the intersection of Multi-Agent Systems, Game Theory, and Cognitive Science, driven by the question: how does intelligence emerge through interaction? I am particularly interested in building agents that model beliefs and intentions, maintain shared context, and reason strategically in long-horizon, imperfect-information environments.

My work spans neuro-symbolic architectures for strategic reasoning, cognitive-inspired learning paradigms, and LLM-based multi-agent communication. I have developed symbolic-LLM reasoning modules for bridge bidding, implemented recursive Theory-of-Mind planners, and studied biologically inspired learning mechanisms to understand how efficient adaptation arises in both artificial and natural systems.

Across these projects, I aim to integrate symbolic structure, cognitive theory, and deep learning to design socially intelligent agents that coordinate, communicate, and learn through interaction.

🎓 Education

Peking University, Beijing, China
B.S. in Artificial Intelligence, Yuanpei College
Sep. 2022 – Jun. 2026 (expected)

🔬 Research Experience

🃏 LLM-based Bidding System for Contract Bridge

Research Assistant | Advisor: Prof. Yaodong Yang
Jun. 2024 – Present

  • Designed a neuro-symbolic bidding framework for bridge, an imperfect-information multi-agent game.
  • Developed an LLM post-training pipeline and a hybrid reasoning module that maps human bidding principles into machine-interpretable constraints.
  • Built a self-play simulation system for iterative policy refinement and strategy validation.

📄 Paper:
PDF

🤝 Fast Adaptation in Mixed-Motive Multi-Agent Games

Research Assistant | Advisor: Dr. Xue Feng
Feb. 2024 – Jun. 2024

  • Implemented Planning with Theory of Mind (PToM) for intention modeling in mixed-motive games.
  • Conducted benchmarking on social dilemma domains to evaluate convergence, adaptation, and robustness.
  • Demonstrated significant improvements in few-shot adaptation compared with RL baselines.

📄 Paper:
PDF

🧠 Frontiers in Computational Neuroscience

Independent Project | Advisor: Prof. Kai Du
Feb. 2025 – Jun. 2025

  • Conducted a systematic review on interpretability of computational cognitive models.
  • Built a simulation framework for studying how presynaptic release probability modulates synaptic plasticity and learning dynamics.
  • Implemented analysis of presynaptic impact on long-term potentiation (LTP) and short-term adaptation.

📄 Final Project:
PDF

💻 Internship Experience

Beijing PKU Yinghua Technology Co., Ltd.
R&D Intern | Jul. 2025 – Present

  • Optimized legal provision alignment using LCS with position-aware penalties.
  • Developed an LLM-based semantic comparison framework for legal text similarity.
  • Integrated LLM reasoning pipelines into existing database search engines.