Shanghai AI Lab · KnowledgeXLab

Xuemeng Yang

I am currently a young researcher in KnowledgeXLab at Shanghai Artificial Intelligence Laboratory. I earned my Master’s degree from Zhejiang University in 2022, where I was a member of the APRIL Lab under the mentorship of Dr. Yong Liu. Prior to this, I completed my bachelor’s degree at Northwestern Polytechnical University.

My current research focuses on AI agent frameworks. Specifically, I am interested in intelligent agents for industrial software operation, tool utilization, and continuous learning mechanisms. My core objective is to drive Agent technology from theoretical models to solving complex real-world problems. Prior to this, my research focused on autonomous driving.

Feel free to drop me emails (xuemengyang96@gmail.com) if you have similar interests on above topics.

Xuemeng Yang · 杨雪梦

Research Focus

Building agents that learn, act, and evolve within closed-loop environments.

I believe that open-loop execution limits an agent's true potential. To thrive in complex, dynamic scenarios, all agents must operate, learn, and evolve within a continuous closed-loop environment. To achieve this, my research focuses on both agent frameworks and world modeling technologies—breaking the boundaries of static learning to drive the realization of lifelong, self-evolving intelligence.

LLM Agent Frameworks Continuous Learning World Modeling Closed-Loop Simulation Industrial Software AI Cognitive Driving Agents

Preprint 2026

Quo Vadis, World Modeling? Towards Interactive World Proxies for Continually Improving Agents

Yang Y, Yang X, Wen L, et al.

arXiv →

ACL 2026 Findings

The Agent's First Day: Benchmarking Learning & Scheduling in Workplace Scenarios

Fu D*, Mei J*, Wu R*, Yang X*, et al.

arXiv →

ACL 2026 Findings

Learning on the Job: An Experience-driven Self-evolving Agent for Long-horizon Tasks (MUSE)

Yang C*, Yang X*, Wen L*, et al.

arXiv →

TMLR 2026

O²-Searcher: A Searching-based Agent for Open-domain Open-ended Question Answering

Mei J, Hu T, Fu D, Wen L, Yang X, et al.

arXiv →

ICCV 2025

DriveArena: A Closed-Loop Generative Simulation Platform for Autonomous Driving

Yang X*, Wen L*, Wei T*, et al.

arXiv →

NeurIPS 2024

LeapAD: Continuously Learning, Adapting, and Improving Autonomous Driving

Mei J, Ma Y, Yang X, et al.

arXiv →

Xuemeng Yang

Building agents that learn, act, and evolve within closed-loop environments.

Selected Publications