Aoxiang Qin

Aoxiang Qin

MSc Student, AI & Robotics

👋 About Me

Hi! I’m Aoxiang Qin (秦奥翔), currently pursuing an MSc in Artificial Intelligence and Robotics at CUHK-Shenzhen.

My research interests include foundation models, value alignment, mechanistic interpretability, efficient long-sequence modeling, and embodied intelligence. I am particularly interested in building interpretable, efficient, and trustworthy AI systems.

Feel free to reach out via email or connect with me on LinkedIn and GitHub.

🔥 News

  • 2026.08: Our paper Steer on Rubrics has been accepted to EMNLP 2026.
  • 2026.08: Our paper Mamba with Hierarchical Memory is available on arXiv.
  • 2026.08: Serving as a student co-organizer for PlurVA-LLM, an AACL 2026 workshop.
  • 2026.04: 🎉 Launched my personal homepage.

📝 Publications and Manuscripts

🔬 Research Experience

Research Student, Freedom AI, CUHK-Shenzhen, Shenzhen, China
Apr. 2025 - Present · Supervised by Prof. Benyou Wang and Dr. Li Zhou

  • Conduct research on value alignment, mechanistic interpretability, and controllable behavior steering for language models and multimodal agents.
  • Build contrastive value-related datasets and trace cross-layer information flow with Contextual Decomposition for Transformers.
  • Localize and causally validate sparse circuits associated with value-related model behavior using activation and attention-head ablations.

Research Assistant, Department of Computer Science, City University of Hong Kong, Hong Kong, China
Jul. 2025 - Apr. 2026 · Supervised by Prof. Zhichao Lu

  • Investigated robust spiking neural networks and neuromorphic computing in a joint research setting with Huawei 2012LAB.
  • Explored efficient, noise-robust training methods for low-latency SNNs and their connection to long-sequence modeling.

💼 Internships

Research Intern, Huawei 2012LAB, Shenzhen, China
Jul. 2025 - Apr. 2026 · Neuromorphic Computing and Spiking Neural Networks · Supervised by Dr. Luziwei Leng

  • Developed a noise-aware ANN-to-SNN conversion and fine-tuning framework for robust, low-latency image classification.
  • Implemented Batch Normalization folding, post-conversion calibration, and LSQ-based quantization-aware fine-tuning for SNN deployment.
  • Extended the framework to long-sequence benchmarks for efficient and robust spiking sequence modeling.

Algorithm Engineer Intern, Baidu Era Technology Co., Ltd., Beijing, China
Jun. 2024 - Sep. 2024 · Speech-Text Multimodal Dialogue Agent

  • Developed a speech-text multimodal dialogue agent based on Baidu’s PANDA LLM, improving question-answering accuracy by 18%.
  • Integrated and optimized Whisper for automatic speech recognition, reducing word error rate to 5.2%.
  • Applied quantization-aware training to BERT models for efficient deployment, reducing model size to 30% of the baseline.

📖 Education

  • MSc in Artificial Intelligence and Robotics, CUHK-Shenzhen, Shenzhen, China
    Sep. 2024 - Present
  • BSc in Computer Science, Ocean University of China, Qingdao, China
    Sep. 2020 - Jun. 2024 · Undergraduate thesis: Research on Underwater Laser Line Enhancement Algorithms · Advisor: Prof. Junyu Dong

🏆 Awards

  • Second Prize for Postgraduate Entrance Scholarship, The Chinese University of Hong Kong, Shenzhen, Sep. 2024.
  • Outstanding Undergraduate Thesis, Ocean University of China, Jun. 2024.
  • Outstanding Student, Ocean University of China, Sep. 2023.
  • First-Class Scholarship, Ocean University of China, Sep. 2023.
  • Lu Xin Social Scholarship, Ocean University of China, Sep. 2023.
  • Second Award (National Level), China Undergraduate Mathematics Competition (Math Major), Nov. 2023.
  • Dean’s Award, Xingyuan College, Ocean University of China, Oct. 2022.
  • Third-Class Scholarship, Ocean University of China, May 2021.
  • Boya Award, Xingyuan College, Ocean University of China, Nov. 2021.

🤝 Academic Service