Welcome to Yun Hong’s Academic Portfolio
About Me
I am Yun Hong (洪运), a Computer Science student at Fudan University (复旦大学) in Shanghai, China, and currently an exchange student at the University of Texas at Austin. My research interests span Machine Learning, Robotics, and Autonomous Driving.
“So we beat on, boats against the current.”
Current Research
Research on ChatGPT-o1 Model Reproduction and Reflection Ability Improvement
Research Assistant | Fudan University Natural Language Processing Group
Sep. 2024 - Jun. 2025
Under the guidance of Professor Xuanjing Huang, I work on introducing self-inspection and correction mechanisms into model reasoning to improve performance in complex tasks. My responsibilities include:
- Synthesis and training of self-critic data based on Llama3.1-8b-instruct
- Generating “reflection” answers from wrong to right using diverse strategies
- Constructing complex reflection paths based on tree search reasoning
- Performance evaluation using GSM8K test sets
Learning Risk Map for Autonomous Driving in Partially Observable Environments
Research Focus | MagicLab, Fudan University
Oct. 2024 - May 2025
Working under the guidance of Wenchao Ding on risk maps for autonomous driving:
- Engineered risk field representations using advanced spatiotemporal modeling techniques
- Developed realistic traffic scene generation leveraging diffusion models with gradient optimization
- Designed a lightweight neural network for efficient risk prediction
- Research submitted to IEEE Robotics and Automation Letters (RAL)
Education
| Fudan University | Shanghai, China |
| B.S. Candidate in Computer Science | Sep. 2023 - Jul. 2027 (Expected) |
| University of Texas at Austin | Austin, Texas |
| Exchange Student in Computer Science | Aug. 2025 - Dec. 2025 |
Technical Skills
- Machine Learning: Completed ML2022 by Hung-yi Lee, proficient in various ML algorithms
- Deep Learning: CNN, RNN, LSTM, Transformer, GAN, Diffusion Models, PyTorch
- Programming: C, C++, Python, Java, LaTeX
- AI-Driven Systems: Autonomous Driving (planning, prediction, decision making)
- Computer Vision: Multi-view Geometry, 3D Reconstruction (SfM/Stereo), Camera Calibration, Object Detection & Segmentation, Optical Flow
- Data Structures & Algorithms: Completed Data Structure (Honor) course by Weiwei Sun
- Languages: Chinese (Native), English (Fluent)
Featured Projects
LCDP-Sim: Language-Conditioned Diffusion Policy
Dec. 2025 - Present | GitHub
- End-to-end Vision-Language-Action (VLA) system utilizing Diffusion Policy
- Maps RGB images and natural language instructions into robotic control signals
- Integrated CLIP text encoder and U-Net-based DDPM/DDIM architecture
MapNavigation
Dec. 2024 - Present | GitHub
- Combines OpenStreetMap data with Gaode Maps API for accurate routing and location search
- Features hybrid point selection, detailed route information, and support for different road types
- Independent development with long-term maintenance commitment
NeuralStyle
Jun. 2025 - Present | GitHub
- Modular neural style transfer toolkit with clear separation for maintainability
- Batch processing capabilities for automated multi-image/multi-style workflows
- Interactive web interface and configurable pipeline
- Open-source with continuous updates
Contact
📧 Email: 23300240019@m.fudan.edu.cn
📍 Location: Shanghai, China
📱 Phone: (+86) 158-0701-0023
💻 GitHub: @16yunH
This website is built using the Academic Pages template and hosted on GitHub Pages.