Portrait
Youwei Liu
Passion for NLP
World Modeling for Agents
About Me

I am an undergraduate at Central South University. I am currently a Research Assistant at The Hong Kong Polytechnic University (PolyU NLP Lab), supervised by Prof. Maggie Wenjie Li.

My research focuses on long-horizon agent planning by integrating LLM agents, embodied agents, reinforcement learning, and world models. I collaborate with Jian Wang on agent research.

My latest CV is available here (Jun. 2026).

Education
  • Central South University
    Data Science Excellence Talent Program
    Undergraduate Student Sep. 2023 - 2027 (expected)
Experience
  • PolyU NLP Lab
    The Hong Kong Polytechnic University
    Research Assistant Apr. 2025 - Present
  • CUHK (Shenzhen) and KAUST
    visiting student Jul. 2024
News
2026
Co-Evolving World Models and Agent Policies (COMAP) was released as an arXiv preprint.
Jun 01
Imagine-then-Plan (ITP) was released as an arXiv preprint.
Jan 01
Selected Publications (view all )
Co-Evolving World Models and Agent Policies for LLM Agents

Youwei Liu, Jian Wang, Hanlin Wang, Wenjie Li

arXiv preprint 2026

A closed-loop framework for co-evolving world models and agent policies, using future-aware reflection to refine agent actions with imagined next-state feedback.

[Paper] [PDF] [Code]

Co-Evolving World Models and Agent Policies for LLM Agents

Youwei Liu, Jian Wang, Hanlin Wang, Wenjie Li

arXiv preprint 2026

A closed-loop framework for co-evolving world models and agent policies, using future-aware reflection to refine agent actions with imagined next-state feedback.

[Paper] [PDF] [Code]

Imagine-then-Plan: Agent Learning from Adaptive Lookahead with World Models

Youwei Liu, Jian Wang, Hanlin Wang, Beichen Guo, Wenjie Li

arXiv preprint 2026

A framework for agent learning via lookahead imagination with a learned world model, including adaptive lookahead for balancing ultimate goals and task progress.

[Paper] [PDF] [Code]

Imagine-then-Plan: Agent Learning from Adaptive Lookahead with World Models

Youwei Liu, Jian Wang, Hanlin Wang, Beichen Guo, Wenjie Li

arXiv preprint 2026

A framework for agent learning via lookahead imagination with a learned world model, including adaptive lookahead for balancing ultimate goals and task progress.

[Paper] [PDF] [Code]

All publications
Honors & Awards
  • Outstanding Paper (Oral), Shenzhen Cup Math Modeling Challenge
    Aug. 2024
  • First Prize, National Undergraduate Math Modeling Competition
    Sep. 2024
  • First Prize, National Undergraduate Market Research Analysis (CP Cup)
    May 2024
  • Second Prize, CSU Mathematical Modeling Competition
    Jun. 2024
  • National-Level Innovation and Entrepreneurship Training Award
    Apr. 2024
  • National Second Prize, Service Outsourcing Innovation and Entrepreneurship Competition
    May 2025
  • Honorable Mention, Mathematical Contest in Modeling (MCM), COMAP-USA
    Jan. 2025