Xiaoqiang Shi

I am a PhD student in Computer Application Technology at the University of Chinese Academy of Sciences. My current interests include multimodal large language models, agent systems, computer vision, and 3D Gaussian Splatting based avatar reconstruction.

My recent work spans two connected directions. On the foundation-model side, I have hands-on experience with decoder-only Transformer training pipelines, tokenizer and data construction, pretraining, SFT/LoRA fine-tuning, reinforcement learning training, sampling, KV cache, multimodal alignment, VQA/caption prototypes, and tool-using coding agents. On the vision side, I work on semantic segmentation, object detection, and high-fidelity animatable human/head avatars based on 3D Gaussian representations.

Before my PhD study, I received my B.Eng. degree in Mechanical Design, Manufacturing and Automation from University of Science and Technology Liaoning. I use this site to maintain my project portfolio, CV, publications, and research/engineering notes.