Research Projects

My research focuses on developing theoretical foundations and practical algorithms for intelligent systems, with a focus on aligning AI with human values and understanding cognitive processes.

Current Projects

AI Value Alignment & Agent Intelligence

2023 - Present
Beijing Institute for General Artificial Intelligence (BIGAI)

Research on building value-aligned autonomous agents, including:

  • Modeling human value systems and their alignment with large language models
  • Developing intrinsic drive mechanisms for LLM-based agents
  • Studying theory of mind in AI agents and its application to cooperative decision-making
Value Alignment LLMs Agents Theory of Mind

Human Learning & Cognitive Modeling

2024 - Present

Investigating computational models of human conceptual learning and cognitive resource limitations, with validation through MEG neuroimaging data.

Cognitive Science Bayesian Inference Neuroimaging

LLM/VLM Interpretability & Training Dynamics

2024 - Present

Analyzing training dynamics of large language and vision models, including curriculum learning, gradient flow, and reward model alignment.

Interpretability Training Dynamics VLMs

Past Projects

Cooperative Communication & Social Intelligence

2019 - 2021
Rutgers University CoDaS Lab
  • Developed mathematical frameworks for cooperative communication as belief transport
  • Studied social intelligence in human-AI teams and built evaluation benchmarks
  • Published work in NeurIPS, ICML, and Trends in Cognitive Sciences
Cooperative Communication Social Intelligence Optimal Transport

Data Compression for HPC

2021 - 2023
  • Designed lossy/lossless compression algorithms for high-performance computing applications
  • Developed preconditioning techniques to improve compression ratio and speed
  • Published in IEEE Transactions on Big Data, SOCC, and IPDPS
Data Compression HPC Algorithms

Game AI & Player Behavior Modeling

2021 - 2022
NetEase Interactive Entertainment
  • Built player strategy mining systems for SLG games
  • Optimized matching algorithms for battle royale game modes
  • Developed AI opponents with adaptive difficulty
Game AI Player Modeling Reinforcement Learning

Research Interests

Machine Learning

  • Theory of Mind
  • Cooperative Game Theory
  • Reinforcement Learning
  • LLM Alignment
  • Bayesian Methods

Mathematics

  • Optimal Transport
  • Algebraic Geometry
  • Bayesian Inference
  • Markov Chains
  • Statistical Inference

Cognitive Science

  • Human Concept Learning
  • Value Modeling
  • Neuro-Symbolic AI
  • Social Intelligence
  • Theory of Mind