Research
Current Research Projects
Reinforcement Learning for Poker
Exploring reinforcement learning approaches for Texas Hold’em poker, focusing on strategy optimization and opponent modeling.
Digital Human Generation
Developing models for digital human skeleton-based content generation, with emphasis on realistic motion and expressiveness.
AI for Education
Building AI systems to support children’s foreign language learning and interactive tutoring.
Financial Intelligence Agents
Designing agents for fraud detection, anomaly mining, and financial market decision support, with multiple papers in preparation.
Stock Market Mining
Researching intelligent methods for stock market analysis and prediction using reinforcement learning and agent-based models.
Multimodal Geo Foundation Models
Developing multimodal models for geospatial understanding, integrating vision, text, and structured data.
Intelligent Customer Service
Building large-scale intelligent customer service models to enhance automated dialogue and query resolution.
Intelligent Transportation Modeling
Exploring multi-agent reinforcement learning approaches for traffic flow optimization and safe driving behavior.
Workflow Optimization
Studying dynamic workflow orchestration and agent routing for complex computational tasks.
Symphony: Decentralized Multi-Agent Framework
Developing a decentralized system where multiple agents collaborate across distributed nodes to solve large-scale problems.
Reward Benchmark
Designing a benchmark for evaluating reward functions in reinforcement learning, focusing on multi-agent environments.
Gradientsys: Multi-Agent Scheduler
Creating a scheduler that coordinates multiple LLM agents with ReAct-based orchestration for efficient task solving.
Self-Evolving Reinforcement Learning
Building reinforcement learning agents capable of continuous self-improvement and adaptive memory evolution.
Memory Evolve
Developing scalable distributed memory systems for agents, enabling long-term learning and coordination.
Computer Use with Meta
Training reinforcement learning agents to operate in web environments through world-model based approaches.
MassGen Collaboration
Working with research partners on generative multi-agent systems for interactive multi-turn reasoning.
Pokémon Agents
Using reinforcement learning in game environments to study self-evolving behaviors and long-term planning.
Workflow Optimization with Multi-Agent RL
Applying multi-agent reinforcement learning to workflow scheduling and dynamic process optimization.
Computer Vision for Surveillance
Developing computer vision methods for public surveillance and safety monitoring applications.
Post-Training Optimization
Improving post-training methods such as GTPO for more effective alignment of large language models.
LLM Red-Teaming
Studying adversarial attacks and defense strategies to enhance the safety and robustness of LLMs.
Medical LLM
Developing large language models specialized for healthcare and medical decision-making.
Virtual Try-On
Researching AI-driven virtual try-on systems with improved realism and personalization.
LLM Reasoning
Enhancing reasoning capabilities in large language models through collaborative experiments and agentic design.
Reinforcement Learning for Architecture
Applying reinforcement learning to building design, optimization, and structural engineering tasks.
AI for Software Engineering
Exploring LLM-based tools and agents for software development and debugging automation.
AI Safety
Investigating safe deployment strategies for AI systems, including red-teaming and adversarial evaluation.
LLM Hardware Acceleration
Researching inference acceleration techniques such as speculative decoding for large-scale models.
IMAE Project
Improving vision-language models with innovative multimodal architectures and training paradigms.
For collaboration opportunities or more information about my research, please contact me.