CHENGJIE FANG / RESEARCH ARCHIVEUCAS 2026.11

06 / Curriculum vitae

Chengjie Fang

Scientific AI · Embodied intelligence · Autonomous research agents

fangchengjie24@mails.ucas.ac.cn
01

Profile

I am a master's student in Artificial Intelligence at the Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences, advised by Prof. Jungang Xu. My research focuses on scientific machine learning, embodied intelligence, and autonomous research agents. I study how physical structure, task representations, and feedback can inform learning systems, with applications in ocean carbon-flux forecasting, long-horizon navigation, and autonomous experimentation. I expect to graduate in June 2027.

02

Education

Sep 2020 — Jun 2024

Wuhan Institute of Technology

Undergraduate study in Intelligent Science and Technology

School of Computer Science and Engineering

Outstanding Graduate. Admitted to UCAS through the graduate recommendation pathway.

03

Research experience

ByteDance · LLM Algorithm Intern

Jul 2026 — Present

TikTok Global E-commerce · Data · Beijing, China

  • Build agents for cross-border e-commerce logistics and supply-chain forecasting, connecting data, experimentation, and modeling platforms to automate information retrieval, task orchestration, and analysis.
  • Apply autonomous research workflows to delivery-time prediction through feature exploration, experiment design, result analysis, and iterative model evaluation.

Baidu · Agent Algorithm Intern

Mar — Jul 2026

BMU Foundation Model Department · Frontier Research · Beijing, China

  • Developed an Auto Research framework for autonomous planning and iterative improvement in complex scientific workflows, with a focus on exploration strategies, context-state management, and sustained reasoning.
  • Designed and implemented the agent framework and experiment harness; contributed to algorithm development, experiments, and system validation.

Wuhan Institute of Technology · Undergraduate robotics research

Undergraduate study

Advisors: Tongzhou Zhao and Xun Li

  • Joined the university's robotics laboratory in my second year and worked on image processing and control for humanoid robots.
  • Led a robotic-arm board-game project in my third year, using deep learning and reinforcement learning. The project was entered in national student competitions.
04

Publications

7 published or accepted papers. * Equal contribution.

  1. Decoupling Regime and Mechanistic Heterogeneity via Hierarchical Mixture-of-Experts for Global Carbon Flux Prediction

    Chengjie Fang, Enyuan Zhao, Hongming Yao, Yujia Yang, Di Gao, Yanbiao Li

    ACM MM 2026 CCF A First author Accepted
  2. StageNav: A Diagnostic Benchmark and State-Aligned RL Framework for Long-Horizon Navigation

    Chengjie Fang, Yujia Yang, Yu Liu, Ziyu Sun, Jungang Xu

    IROS 2026 CCF C First author Accepted
  3. Carbon-ECS: A Benchmark and a Physics-Decoupled Model for the East China Sea Carbon Flux

    Chengjie Fang, Enyuan Zhao, Hongming Yao, Ziyu Sun, Di Gao, Yanbiao Li

    ICASSP 2026 CCF B First author Published
  4. ForesightNav: Adaptive Semantic-Geometric Planning for Universal Zero-Shot Goal-Oriented Navigation

    Yu Liu*, Chengjie Fang*, Likun Lu

    IEEE Access 2026 IF 4.2 · 2025 Co-first author · 2nd listed Published
  5. SHNet: A Lightweight Network Focusing on Shallow High-Resolution Features for PCB Defect Detection

    Yu Liu*, Chengjie Fang*, Likun Lu

    IEEE Access 2026 IF 4.2 · 2025 Co-first author · 2nd listed Published
  6. Filter-Group Mixture-of-Experts Model for Remote Sensing Forged Target Perception

    Ziyu Sun*, Enyuan Zhao*, Chengjie Fang, Wei Xiong, Heng Pan, Yanbiao Li

    ICASSP 2026 CCF B Third author Published
  7. AnyTemplate: A Zero-shot Template Matching Approach for Arbitrary-Resolution and Scale-Agnostic Images

    Dongjie Chen, Zhiwei Lv, Bowen Liu, Chengjie Fang, Jungang Xu

    IJCNN 2026 CCF C Fourth author Published
05

Manuscripts under review

Research agents

UNDER REVIEWFirst author

What Models Miss: Scientific State Representations for Autonomous PDE Modeling

Scientific state representations and residual-driven exploration for autonomous PDE model search and refinement.

Scientific AI

UNDER REVIEWFirst author

Decoupling Temporal Reaction Search from Potential-Wave Spatial Response

Separating temporal reaction search from differentiable spatial physics for ocean carbon-flux forecasting.

06

Selected honors & awards

Graduate study

National Scholarship

University of Chinese Academy of Sciences

2026

Third Prize · Beijing Challenge Cup

BioClaw: scientific literature mining for life-science research resource matching

Nov 2025

National First Prize · China Robot Competition & RoboCup China Open

Team Captain

Jun 2025

National Second Prize · China Robot and Artificial Intelligence Competition

Innovation Track · Team Captain

Jun 2025

National Second Prize · 16th Lanqiao Cup

Python Programming · National Finals

Undergraduate study

Outstanding Graduate

Wuhan Institute of Technology

07

Technical & research skills

Robotics & embodied AI

ROS on Linux, Gazebo, Isaac Gym, SLAM, robot navigation, and physical-system debugging.

Learning & agents

Reinforcement learning, computer vision, natural language processing, scientific machine learning, agent development, and LLM fine-tuning.

Research practice

Problem formulation, experimental design, paper reproduction, literature review, and scientific writing.

Research figure