🔍 About Me

I am Qihang Yang, an M.Sc. student in Computer Science at the University of Hong Kong. Previously, I was a Research Assistant at the Institute of Information Engineering, Chinese Academy of Sciences, where I worked on reliable code generation and LLM post-training. At UESTC, I conducted research on uncertainty-aware multimodal perception for autonomous driving.

My research interests include trustworthy AI, multimodal learning, computer vision, uncertainty estimation, and large language models. I am particularly interested in building trustworthy learning systems through principled modeling and rigorous empirical evaluation.

🔬 Research Experience

MineValiCoder: Reliable Code Generation with Test Case Quality Mining

Research Assistant, Institute of Information Engineering, Chinese Academy of Sciences · Beijing, China · 2025.07–2026.08

• Developed a closed-loop test-driven development framework that improves LLM-generated code without relying on human-written test cases.
• Designed self-validating test generation and dynamic code–test mutual scoring, and evaluated the framework on HumanEval, MBPP, and APPS using GPT-4, Llama-3.1-8B, Qwen2.5-Coder-7B, and Qwen3-4B. [arXiv]

Uncertainty-Aware Multimodal Perception for Autonomous Driving

Research Assistant, Center for Robotics, UESTC · Chengdu, China · 2024.02–2025.05

• Developed a plug-and-play late-fusion framework for fusing camera and LiDAR detections while preserving interpretable predictive uncertainty.
• Integrated evidence-theoretic uncertainty quantification into classification fusion using subjective logic and Dempster–Shafer theory, reducing predictive uncertainty by up to 78% on KITTI; results were published in Drones and at CVCI 2025. [Drones] [CVCI 2025]

PETRFusion: Multisensor BEV Semantic Segmentation

Undergraduate Research Assistant, Center for Robotics, UESTC · Chengdu, China · 2023.03–2024.05

• Developed the image branch and improved the 3D positional encoder through depth estimation and residual structures.
• Conducted ablation studies on image backbones, encoder designs, and data-pipeline configurations. [ITSC 2024]

📃 Publications

• MineValiCoder: Reliable Code Generation with Test Case Quality Mining and Bipartite Graph-Based Mutual Validation, Z. Zhao, Qihang Yang, F. Dai, X. Li, B. Li, arXiv preprint, 2026

• Uncertainty-Aware Evidential Fusion for Multi-Modal Object Detection in Autonomous Driving, Qihang Yang, Yang Zhao, Hong Cheng, Drones, 10(2):130, 2026

• MMLF: Multi-modal Multi-class Late Fusion for Object Detection with Uncertainty Estimation, Qihang Yang, Yang Zhao, Hong Cheng, 2025 IEEE 9th CAA International Conference on Vehicular Control and Intelligence (CVCI)

• PETRFusion: Multi-Sensor Fusion Based BEV Semantic Segmentation Network for Autonomous Driving, Yang Zhao, Jingyu Du, Qihang Yang, Ningze Cai, Zhinan Peng, Huiqin Zhan, Hong Cheng, 2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC)

🎓 Education

• 2026–2028 (expected), M.Sc. in Computer Science, The University of Hong Kong, School of Computing and Data Science, Shanghai, China
• 2020.09 - 2024.06, B.Eng in Electronic Engineering, University of Electronic Science and Technology of China (Glasgow College), Chengdu, China — GPA 3.53/4.00

💻 Technical Skills

• Programming: Python, C, MATLAB, Bash
• Deep Learning Frameworks: PyTorch, Verl
• LLM Post-Training: SFT, DPO, GRPO; rollout, reward computation, and training and evaluation pipelines
• Multimodal Perception: Object detection, BEV segmentation, camera–LiDAR fusion, and uncertainty estimation; YOLOv8, DETR, BEVFusion, and deep ensembles
• Development Tools: Git, Linux, LaTeX

⭐ Honors

• 2022.12 UESTC Standard Student Scholarship