Research

My Research Vision

Driven by the wave of digitalization and intelligentization, artificial intelligence is profoundly transforming how humans interact with the world. My research vision focuses on multi-agent systems, human-machine collaboration, decision intelligence, and multimodal interaction to explore innovations in complex systems empowered by AI, promoting deep integration of agent technologies in both academic and real-world scenarios.

My focus extends beyond improving algorithmic performance—I aim to develop a new generation of intelligent systems throughout the entire "intelligence-cognition-interaction-trust" chain that are explainable, scalable, and capable of self-reflection and autonomous collaboration. Whether working on digital twins, intelligent storytelling, AI-assisted optimization, financial risk analysis, or general multimodal generation engines, I strive for original technological innovations to break through the limitations of fragmented and black-box AI, laying the foundation for an open, trustworthy, and practical AI-powered society.

Moving forward, I look forward to collaborating with interdisciplinary partners to continuously expand the intersection of AI with society, economics, cognition, and humanities—together creating a smarter world that's more humane and imaginative.

RTCO Decision-Making

Round-Table Consenseus Operation Curriculum with Multi-Agent: The RTCO project is dedicated to building an automated optimization and verification framework centered around multi-agent collaboration. We incorporate a "roundtable consensus" mechanism into the modeling, solving, and self-verification of optimization problems, enabling expert role allocation, automated reasoning, and reflective correction. Our goal is to create highly reliable and interpretable intelligent optimization systems for combinatorial optimization, reasoning verification, complex decision-making, and related fields. RTCO focuses not just on algorithms themselves, but emphasizes multi-agent debate, consensus-building, and transparency - introducing a new paradigm for AI optimization.

Project AENOVA

Advanced Embodied Neural Omni-modal Virtual Assistant: AENOVA aims to explore the possibilities of next-generation "digital twins" and human-machine symbiotic intelligence. We pursue not just natural language interaction, but intelligent systems that integrate multimodal perception (vision, speech, motion), autonomous decision-making, and contextual memory. AENOVA's vision is to empower users with entirely new ways of information interaction, enabling virtual assistants to truly understand, perceive, and proactively serve in complex, ever-changing real-world scenarios. Moving forward, we will continue to drive AI innovation in cutting-edge areas like emotional understanding, barrier-free communication, and multi-device collaboration.

SYNAPSE System

Synthetic Agents for Pluralistic Strategic Engagement: SYNAPSE represents a novel multi-agent architecture for complex reasoning and strategic gameplay. We focus on constructing diverse, heterogeneous AI agent collectives capable of collaboratively solving complex tasks through knowledge sharing, role differentiation, and intelligent gaming. SYNAPSE's vision is to empower AI collectives with higher-level strategic behaviors and creativity, driving systematic breakthroughs from distributed reasoning and cooperative decision-making to adaptive strategy generation. This project will provide the foundation for intelligent gaming, AI collaborative innovation, and the implementation of new Agentic Frameworks.

ORACULUM Fin

Optimized Risk Assessment and Computational Understanding of Latent Uncertainty in Markets: ORACULUM is our flagship platform for the future of FinTech, focusing on leveraging AI and machine learning methods to deeply understand and predict financial market uncertainties and risks. The project covers multiple areas, including volatility modeling, explainability analysis, transfer learning, adaptive high-frequency trading, sentiment analysis, and macro-factor integration. ORACULUM's vision is to build a transparent, verifiable financial AI engine that can assist in decision-making, providing innovative tools for investors, regulators, and scholars, and empowering the financial industry with intelligence and upgraded risk management.

LUMINA Game AI

Logic-Underpinned Modular Interaction for Narrative Agents: The LUMINA project focuses on creating next-generation narrative interaction systems for games and virtual worlds with minimal fragmentation and strong information consistency. We explore multi-agent based story structure modeling, beat logic verification, and dynamic completion methods to ensure natural, fluid, and profound dialogue and event progression. LUMINA aims to break through the limitations of traditional interactive experiences, making every conversation and choice between players and virtual worlds logically coherent and continuous. We believe this will establish "intelligent storytelling" as the new standard for future digital entertainment and interactive media.

Project ALMA

Advanced Large Model Architecture Exploration: ALMA is dedicated to exploring next-generation multimodal generation and interaction models, encompassing unified modeling and intelligent synthesis of diverse data types including text, images, audio, and motion. We focus on tokenization innovations, diffusion/autoregressive/hybrid generation paradigms, and novel unified inference and execution architectures. ALMA aims to break through modality barriers to achieve more natural, intelligent, and efficient human-computer interaction and content creation. Our vision is to build an adaptive, highly scalable general AI engine that drives the development of future intelligent applications and multimodal systems.

Page updated

Google Sites

Report abuse