Decision-Making
Large language models are becoming decision interfaces for complex systems, but reliable decisions still require domain grounding, uncertainty awareness, and actionable intervention design. We develop decision-making methods that use LLMs to connect data, language, and optimization, enabling applications such as public health analysis, personalized decision modeling, and trustworthy traffic safety assessment.
Agentic AI and Collaborative Intelligence
LLMs are powerful collaborators, but their reasoning becomes more reliable when it is shaped by human priors, physical constraints, and domain knowledge. Our work on collaborative intelligence studies how LLMs can organize, evaluate, and improve their own reasoning through self-governance, and how LLM-based mixture-of-experts systems can coordinate specialized agents to produce grounded, cooperative intelligence.
Computing and Intelligence Scaling
Scaling LLMs and multimodal foundation models is not only about making models larger; it also requires efficient computation, adaptive routing, and deployable intelligence under real-world resource constraints. We design computing and intelligence scaling methods for energy and transportation systems, edge AI, computing efficiency, and LLM life assessment that bring large-model intelligence closer to practical deployment.
Advancing real-time infectious disease forecasting using large language models
Nature Computational Science
Personalized Decision Modeling: Utility Optimization or Textualized-Symbolic Reasoning
Advances in Neural Information Processing Systems (NeurIPS), 2025 Spotlight
How to Auto-optimize Prompts for Domain Tasks? Adaptive Prompting and Reasoning through Evolutionary Domain Knowledge Adaptation
Advances in Neural Information Processing Systems (NeurIPS), 2025
SafeTraffic Copilot: adapting large language models for trustworthy traffic safety assessments and decision interventions
Nature Communications volume 16, Article number: 8846 (2025)
Toward Optimal Mixture of Experts System for 3D Object Detection: A Game of Accuracy, Efficiency and Adaptivity
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)
Towards Accurate and Efficient 3D Object Detection for Autonomous Driving: A Mixture of Experts Computing System on Edge
International Conference on Computer Vision (ICCV), Poster
Towards Physics-informed Spatial Intelligence with Human Priors: An Autonomous Driving Pilot Study
Advances in Neural Information Processing Systems (NeurIPS), 2025 Spotlight
Cost-effective vehicle recognition system in challenging environment empowered by micro-pulse lidar and edge AI
2024 IEEE Intelligent Vehicles Symposium (IV)