Research Interests
- AI/LLMs for Simulation Optimization
- Stochastic Optimization
- Supply Chain Management
- Service Operations
- Reinforcement Learning
Direction I: LLMs × Simulation + Optimization
(JMP) Yanyuan Wang and Xiaowei Zhang (2026). Optimizing Service Operations via LLM-powered Multi-agent Simulation. Paper (Poster).
- Second Prize, 2026 POMS-China Best Student Paper Competition
TL;DR Integrate LLMs into system design to enable behavioral awareness through a human-AI collaboration paradigm.
- Human: system designer
- AI: LLM-powered multi-agent system (in-silicon samples)
LLM-MAS functions as both an explorer and an evaluator of system designs, serving as a bridge that connects game-theoretic analysis with real-world dynamics.
Envisioning Future Research Directions in LLM-MAS&O (LLM-powered Multi-Agent Simulation & Optimization)
- Theoretical-depth: refined algorithm (work in progress)
- High-fidelity: ensure behaviorally aligned systems (post-training, model choice) (work in progress)
- Robustness: address configurations uncertainty, adapt to evolving environments, and control delusion
- Efficiency: achieve computational speedups
- Scalability: scale up to large systems
- Applicability: enable domain-specific applications
- strategic queueing systems
- online marketplaces (e.g., gig-economy platforms)
- blockchain-based auctions
- cloud computing management
- (...other mechanism-design problems in dynamic markets)
📢 I'm always happy to connect, exchange ideas, and explore collaborations!
Direction II: ML/GenAI × Optimization
Yanyuan Wang and Xiaowei Zhang (2026). “Over-optimizing” for Normality: Budget-constrained Uncertainty Quantification for Contextual Decision-making. Major Revision at Manufacturing & Service Operations Management.
TL;DR Inspired by the AI triad: computing power, data and algorithm. Examine the statistical-computational tradeoff in quantifying uncertainty for context-aware decision-making, revealing a counterintuitive answer to “When does more data stop helping?”
Direction III: ML × Simulation
Wenjia Wang, Yanyuan Wang and Xiaowei Zhang (2024). Smooth Nested Simulation: Bridging Cubic and Square Root Convergence Rates in High Dimensions. Management Science.
TL;DR Introduce a sample-efficient approach to high-dimensional uncertainty quantification by framing unknown function estimation as a machine learning task and leveraging smoothness to reduce functional search complexity. Applications include risk management and uncertainty propagation in stochastic systems across healthcare operations, biopharmaceutical manufacturing, and power grid scheduling, etc.
