ZHANG Motong (张沫潼)

Senior Undergraduate at Xidian University · Electronic Information Engineering

🏁 Actively seeking Ph.D./Research Assistant positions in World Model & Embodied AI for Fall 2026 or later!

📧 karinazhang0518@outlook.com 🔗 GitHub 🎓 Google Scholar 💼 LinkedIn

About Me

I'm a senior undergraduate at Xidian University, majoring in Electronic Information Engineering. My research journey started with computer vision and multimodal perception — I worked with Dr. Hao Li on adversarial robustness through reinforcement learning, with Prof. Rui Yang on ML-driven optimization for RF/microwave systems, and with Prof. Jie Li on RGB-Thermal multimodal object tracking, where I explored how to adaptively fuse information from different sensory modalities under varying conditions.

Currently, I'm interning at Li Auto, working on code-agent architectures and LLM post-training, which has given me a deeper understanding of how large models learn to reason, plan, and interact with environments. Looking ahead, I'm especially excited about World Models and Embodied AI — building systems that can not only perceive and reason about the physical world, but also take actions within it. I see a natural progression from multimodal perception to agentic reasoning to physical intelligence, and that's the trajectory I want to pursue in my Ph.D. research.

I scored IELTS 8.0 (Listening 9.0, Speaking 7.0), and I'm comfortable working and communicating in English-speaking research environments.

Huge THANKS to Alma and Claude for helping me build this website and making differences to the world!

Currently organizing and uploading the code and related files step by step(^-^)

Education

Xidian University, Xi'an, Shaanxi, China Sep. 2022 – Jul. 2026 (Expected)
B.Eng. in Electronic Information Engineering
GPA: 3.6 / 4.0  |  Ranking: Top 15%
Scholarships: Individual Scholarship, Second-class University Scholarship, Linkidge Capital Scholarship
Core Courses: Advanced Programming (Python): 98 (A+), Computer and Programming (C): 91 (A), Data Structure and Algorithm Application: 86 (A-)
Language: CET-6: 666/710  |  IELTS: 8.0 (R 9.0, L 9.0, S 7.0, W 6.5)

Technical Skills

💻 Programming & Systems & Tools
Python C/C++ MATLAB LaTeX CUDA Linux FPGA PCB RISC-V
🧠 Deep Learning & Vision
PyTorch CNN Transformers YOLO GAN LSTM Contrastive Learning Multimodal Fusion ResNet Swin Transformer Object Tracking Image Segmentation
🤖 LLM & Agent Systems
SFT RLHF/PPO Data Pipeline Code Agent Reward Modeling Agent Architecture Synthetic Data
📊 Math & Data Analysis
Optimization Statistical Modeling Time Series Genetic Algorithm Multi-Objective Optimization Simulated Annealing Markov Chain Linear Programming Monte Carlo

Research Experience

Adversarial Attack via Reinforcement Learning Dec. 2023 – Nov. 2024
Dept. of Electronic Engineering, Xidian University
Research Assistant & Project Leader, advised by Dr. Hao Li (Associate Professor)
  • Developed a novel background adversarial attack method through Proximal Policy Optimization (PPO), improving model robustness against real-world perturbations.
  • Fabricated PCB boards for physical-domain experiments, verifying the effectiveness of adversarial samples in real-world settings.
  • Led the student research team as project leader; project rated Excellent (Top 7%) and selected to the National Innovation and Entrepreneurship Program.
Intelligent Microwave Filter Design May 2025 – Jul. 2025
Dept. of Electronic Engineering, Xidian University
Research Assistant, advised by Prof. Rui Yang
  • Conducted research on intelligent microwave filter design integrating electromagnetic theory with machine learning and multi-objective optimization.
  • Built a 2000-sample design database to ensure comprehensive coverage of the design space.
  • Proposed an intelligent co-optimization mechanism between the ML model and optimization algorithm, enabling efficient exploration of Pareto-optimal filter configurations.
RGB-T Multimodal Object Tracking Nov. 2025 – May 2026 (Expected)
Dept. of Electronic Engineering, Xidian University
Undergraduate Researcher, advised by Prof. Jie Li
  • Reproduced the SOTA RGB-T tracker FMTrack (ViT-Base dual-stream architecture with frequency-aware interaction and multi-expert fusion) on RGBT234 and LasHeR benchmarks, establishing a reliable performance baseline.
  • Identified a systematic over-confidence issue in FMTrack's multi-expert fusion module (MEFM), where end-to-end training causes semantic misalignment in modal reliability prediction.
  • Proposed MQRC, a post-hoc confidence calibration method using independent temperature scaling layers to correct modal quality estimates, enabling more accurate adaptive fusion under varying RGB-Thermal conditions.

Internship Experience

LLM Intern Mar. 2026 – Aug. 2026 (Expected)
Li Auto Inc.
Reveal after my last day
  • Surveyed open-source code-agent datasets and benchmarks (SWE-Gym, Scale-SWE, SWE-Rebench, etc.), conducting cross-dataset overlap analysis and multi-model behavioral comparison to identify optimal data strategies for code-agent training.
  • Built end-to-end data processing pipelines including trajectory-level ID matching, quality filtering, and field-level stitching across heterogeneous data sources, delivering curated training data for the team's SFT pipeline.
  • Contributed to post-training experiments through SFT and RLHF for code-specialized models, evaluating different data compositions and reward modeling strategies to improve agent performance on software engineering tasks.

Projects

Deep Learning Land Classification for Remote Sensing 2025

Applied contrastive learning with an improved ResNet-18 architecture for multispectral satellite imagery classification. Implemented adaptive image enhancement and multi-scale feature fusion, achieving a +32% accuracy improvement over the baseline model.

Python PyTorch ResNet Remote Sensing
Super-Resolution with SRGAN 2025

Designed a residual-based generator-discriminator architecture with sub-pixel convolution for 4× image upscaling. Leveraged VGG-19 perceptual loss to preserve texture realism, producing visually sharper and more detailed super-resolved images.

Python PyTorch SRGAN Computer Vision
Quantitative Investment Factor System 2024

Built a text-mining quantitative factor library from financial news, extracting sentiment and event-driven signals. Employed entropy weighting and correlation analysis to construct a robust backtesting framework for factor evaluation.

Python NLP Finance SPSS
Multimodal Medical Image Segmentation 2023

Implemented a Swin-Unet framework for MRI/CT medical image segmentation, achieving 94% accuracy. Applied CUDA acceleration to optimize inference, resulting in a 2.5× speedup in processing time.

Python PyTorch Swin-Unet Medical AI
Financial Vulnerability Detection 2023

Developed a GA-optimized SVM combined with K-means clustering for student financial risk detection. Used PCA to reduce 17 behavioral indicators, achieving 76% accuracy on a dataset of 4,000+ students.

Python SVM K-Means PCA

Awards & Honors

Extracurricular Activities

📣
Core Member, University Publicity Department Oct. 2022 – Sep. 2023

Contributed to university-level media and promotional content creation, helping shape the public image and outreach of the department through articles, graphics, and event coverage.

🎓
University Peer Mentor, Dept. of Electronic Engineering Sep. 2024 – Present

Mentoring junior students in academic planning and course guidance, sharing insights on research opportunities, study strategies, and navigating the engineering curriculum.