Xiaoyu Tao (陶小玉)

Hi! This is Xiaoyu Tao, a PhD student in Computer Science at University of Science and Technology of China (USTC). I am from the State Key Laboratory of Cognitive Intelligence. I am advised by Prof. Qi Liu and currently working with Associate Researcher Mingyue Cheng.

Research Interest

My research focuses on time series analysis and large language models, with particular emphasis on:

• Time Series Forecasting - Developing LLM-driven frameworks for time series forecasting, including memory-driven reasoning, symbolic discretization, and multimodal approaches
• Table Reasoning - Building autonomous programmatic agents for tool-augmented table reasoning with slow-thinking LLMs
• Personalized Generation - Designing hierarchical memory mechanisms from textual interactive behaviors for personalized content generation
• Deep Research Agents - Developing cognitive agents for expert-level commercial report synthesis

My research focuses on time series forecasting and LLM applications. I have published papers on memory-driven time series forecasting (MemCast), LLM-driven context-aware forecasting (TokenCast), slow-thinking LLMs for temporal reasoning (TimeReasoner, WSDM 2026), self-supervised time series representations (TimeMAE, WSDM 2026), hierarchical memory for personalized generation (MemWeaver, WWW 2026), autonomous table reasoning agents (TableMind, STaR), and deep research agents for commercial report synthesis (Mind2Report). I have published 8 papers accepted at CCF A/B tier venues, including 3 as first author or student first author. Additionally, I have 3 papers under review as first author or student first author.

Awards

Experiences