Weicheng Zheng

Weicheng Zheng

Incoming PhD Student

Tsinghua University

Research Interests

Autonomous Driving
Embodied Intelligence
Multimodal Large Language Models

About

I am an incoming PhD student at the Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, where I will join the MARS Lab in September 2026 under the supervision of Prof. Hang Zhao.

I received my B.Eng. in Software Engineering from the School of Computer Science and Technology at Tongji University in June 2026. My research interests include autonomous driving, embodied intelligence, and multimodal large language models.

I am particularly interested in exploring how the generalization capabilities of large language models can be transferred to the physical world.

Research Experience

  • Shanghai Qi Zhi Institute — Research Intern, March 2025–Present
  • Li Auto — Research Intern in Autonomous Driving, April 2025–February 2026

News

2026-02

DriveAgent-R1 was accepted to ICLR 2026 🎉

Publications

DriveMA: Driving Vision-Language-Action Models with Verifiable Meta-Actions

DriveMA: Driving Vision-Language-Action Models with Verifiable Meta-Actions

Weicheng Zheng, Yixin Huang, Qiao Sun, Derun Li, Hang Zhao

arXiv preprint arXiv:2605.31271 2026

A driving vision-language-action framework that uses verifiable meta-actions to close the gap between language-level intent and trajectory planning.

ProjectDOICode
Is Your Trajectory Displacement Safe in Long-tail?

Is Your Trajectory Displacement Safe in Long-tail?

Qiao Sun, Weicheng Zheng, Yixin Huang, Hang Zhao

arXiv preprint arXiv:2606.16313 2026

A human-aligned and safety-aware evaluation framework for identifying additional threats in long-tail autonomous-driving scenarios.

ProjectDOICode
SLAM-Former: Putting SLAM into One Transformer

SLAM-Former: Putting SLAM into One Transformer

Yijun Yuan, Zhuoguang Chen, Kenan Li, Weibang Wang, Minghui Qin, Zhijian Fang, Weicheng Zheng, Hang Zhao

European Conference on Computer Vision (ECCV) 2026

A unified transformer-based SLAM system whose cooperating frontend and backend deliver online tracking, dense mapping, and global refinement.

ProjectDOICode
DriveAgent-R1: Advancing VLM-based Autonomous Driving with Active Perception and Hybrid Thinking

DriveAgent-R1: Advancing VLM-based Autonomous Driving with Active Perception and Hybrid Thinking

Weicheng Zheng, Xiaofei Mao, Nanfei Ye, Pengxiang Li, Kun Zhan, Xianpeng Lang, Hang Zhao

International Conference on Learning Representations (ICLR) 2026

A VLM-based driving agent that improves planning reliability through active perception and adaptive hybrid reasoning.

ProjectDOICode

Awards

Honors & Scholarships

  • Shanghai Outstanding Graduate — June 2026
  • National Scholarship — December 2024
  • Shanghai Scholarship — 2022 & 2023

Academic Service

Peer-review service for the research community.

Conference Reviewer

CVPR · CoRL · ICME

Reviewed research submissions in computer vision, robot learning, and multimedia.