Research Interests
My research focuses on time series and tabular data mining and large language model (LLM) reasoning and agents, with the main goal of building context-aware intelligent decision-making systems to support AI for Science and Recommender Systems.
- Time Series and Tabular Data Mining: We develop core methods for time series and tabular data, focusing on forecasting, classification, representation learning, and multimodal reasoning. Our interests include context-aware representations, non-stationary modeling, and the development of foundation models for universal structured data representation.
- LLM Reasoning and Agent: We aim to enhance large language models via training-free contextual learning and reinforcement learning, focusing on long CoT reasoning, planning, long-context memory, tool use, self-reflection, and multi-agent collaboration. Our objective is to build agents that can perceive, reason, and act autonomously in complex environments.
- Artificial Intelligence for Science: We focus on two directions: (1) understanding structured scientific data, such as time series and tabular data; and (2) building intelligent agents for scientific reasoning, with abilities in inference, planning, and tool use. Our goal is to accelerate discovery by combining data mining and LLM-driven reasoning within autonomous agent systems.
- Recommender Systems: Advance personalized recommendation through sequential and spatio-temporal modeling and one-class collaborative filtering. We investigate reinforcement learning-based methods and leverage large language models to enable context-aware recommendation agents. Our goal is to build generalizable frameworks for dynamic, real-world environments.
欢迎脚踏实地而又积极主动的本科生、研究生同学加入实验室USTC-AGI研究组。欢迎体验我们精心打造的科言智能助写平台提升写作效率与技能。
