Zhiqi Chen

Zhiqi Chen

Undergraduate student at Tsinghua University, studying reinforcement learning for reasoning and agentic robotics.

Research focus: Reinforcement learning, reasoning in language models, agentic robotics.

Latest work: Code as Primitives explores robotic skill synthesis with frontier models and simulation feedback.

Collaboration: Research with Prof. Gao Huang at LeapLab, and research with Haiwen Feng and Prof. Andrea Zanette at Impossible Research and Carnegie Mellon University.

About Me

I am a B.Eng. student in Automation at Tsinghua University, advised by Prof. Gao Huang in LeapLab. My research spans reinforcement learning for reasoning in language models and agentic learning for robotic control. My earlier work examined the limits of reinforcement learning with verifiable rewards (RLVR) in extending LLM reasoning. In Code as Primitives, we study how frontier models synthesize executable low-level robotic controllers and refine them through simulation feedback.

Awards

2026
Sensetime Scholarship
2025 & 2024
National Scholarship of China

Interests

Reinforcement Learning & Reasoning in LLMs and MLLMs,
Agentic Robotics & World Models

Goal

Find the essence of reinforcement learning for reasoning and build a universal theory for reasoning.

Education

Tsinghua University

B.Eng. in Automation · 2023 — present

GPA 4.0/4.0 (Rank 1/173).

Intern

Impossible Inc · 2026 — present

Exploring next-era world models.

Research Intern

LeapLab · 2025 — present

Core contributor to the program Limit-of-RLVR.

Selected Work

Publications

Code as Primitives: Harnessing Frontier Models to Synthesize Low-Level Robotic Skills

Zhiqi Chen*†, Djengo Cyun-Jyun Fang*, Guanning Zeng, Xiuyu Li, Haiwen Feng, Andrea Zanette
Blog
*Equal contribution · †Project lead

Project

Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?

Yang Yue*†, Zhiqi Chen*, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang
NeurIPS 2025 Best Paper Runner-Up Award · ICML 2025 Workshop AI4Math Best Paper Award
*Equal contribution · †Project lead

PDF arXiv Code

Learning Push-Grasp Synergy for Occluded Objects in Cluttered Environments

Ziang Li, Haorui Wu, Zhiqi Chen, Haozhe Zhang, Yuzhe Huang, Changshui Zhang
2026 IEEE International Conference on Robotics and Automation (ICRA)
Vienna, Austria · pp. 1553–1560

IEEE Xplore DOI