👋 About me (Update on 06.01)

I am Songyang Gao (高颂扬), a first-year MPhil student in Artificial Intelligence at The Hong Kong University of Science and Technology (Guangzhou), advised by Prof. Hui Xiong (IEEE / ACM / AAAI / AAAS Fellow). I received my B.Eng. in Artificial Intelligence from Nanchang University in 2025.

My current research interests focus on Agentic Systems and Streaming Systems. I am designing Agentic Systems for Large-Scale Data Tasks, including building a Streaming Agent Benchmark for Real-Time Data Pipelines. In addition, I am also interested in applications of Agentic Systems, such as intelligent agents for scientific research and computer use agents.

Previously, I worked on generative AI algorithms and their applications in computational imaging and intelligent medical imaging at LIVE Lab, Nanchang University, advised by Prof. Qiegen Liu and Wenbo Wan.

📫 I am actively looking for Fall 2027 Ph.D. programs. Please feel free to contact me!

Click here to visit my CV, and click here to read about my personal growth and life story.

Email: sgao068@connect.hkust-gz.edu.cn

🔥 News

• 2026.05:  📄 First-author paper “Graphs of Research” (GoR) is now available on arXiv.
• 2026.02:  📝 SAB (co-first author) is under review at ICDE 2027.
• 2025.12:  🛡️ Served as group leader in the Singtel PIE Program (Singapore), where we proposed SmartShield, a multimodal anti-fraud system.
• 2025.09:  🎓 Started my MPhil at HKUST(GZ), advised by Prof. Hui Xiong.
• 2025.06:  🏫 Graduated from Nanchang University; my undergraduate thesis was rated Excellent.
• 2024.11:  🏆 Received the Baosteel Scholarship Special Prize (only 15 nationwide). Click here to read the official report.
• 2024.08:  📄 Our paper “DMEDH” was accepted by Optics Express.
• 2024.05:  🥇 Won the Best Poster Award (first author) at the 6th Symposium on Computational Imaging.
• 2024.05:  ✍️ Selected as an Outstanding Academic Correspondent for CAAI Transactions on Intelligence Technology.
• 2024.04:  📄 Our paper “HoloDiffusion” was accepted by Photonics.
• 2024.01:  🔬 Attended the Tsinghua AIR winter research camp, advised by Yan Wang.
• 2024.01:  📄 Our paper “Imaging through scattering media” was accepted by Applied Physics Letters.
• 2023.12:  🏅 Received the Huawei “Smart Base” Scholarship (top 0.3%).
• 2023.04:  📄 Our paper “ARA-Net” was accepted by Frontiers in Neuroscience.

📝 Publications

arXiv 2026

Graphs of Research: Citation Evolution Graphs as Supervision for Research Idea Generation

Songyang Gao*, Yinghui Xia*, Siyi Liu, Hui Xiong (*equal contribution) · First Author

arXiv

• We organize the 2-hop citation neighborhood of each paper into a paper-evolution DAG and use it as supervision for LLM-based research idea generation. GoR-SFT attains state-of-the-art performance against gpt-4o-driven baselines in head-to-head LLM-judge tournaments.

ICDE 2027

SAB: A Streaming Agent Benchmark for Real-Time Data Pipelines (Under Review at ICDE 2027)

Yinghui Xia*, Songyang Gao*, Hui Xiong (*equal contribution) · Co-first Author

• We propose SAB, the first systematic benchmark that transforms text-to-SQL tasks into streaming control problems for evaluating LLM-based controllers in production stream processing.

Optics Express 2024

DMEDH: Diffusion Model-boosted Multiplane Extrapolation for Digital Holographic Reconstruction

Songyang Gao, Wei Xu, Xusheng Wu, Ji Liu, Bo Wang, Tong Wu, Wenbo Wan, Qiegen Liu · First Author · (CAS Tier 2)

Paper

• A diffusion-model-boosted multiplane-extrapolation method that improves the quality of digital holographic reconstruction while reducing the cost of optical devices.

Photonics 2024

HoloDiffusion: Sparse Digital Holographic Reconstruction via Diffusion Modeling

Leihong Zhang*, Songyang Gao*, Mengfan Tong, Yiqiang Huang, Zhisheng Zhang, Wenbo Wan, Qiegen Liu (*equal contribution) · Co-first Author

Paper

• A score-based diffusion model combined with physical constraints that enhances digital holographic reconstruction under extremely sparse sensor arrays.

• Imaging through scattering media via generative diffusion model, Zhicheng Chen, Bin Lin, Songyang Gao, Wenbo Wan, Qiegen Liu, Applied Physics Letters, 2024.
• ARA-Net: An Attention-Aware Retinal Atrophy Segmentation Network Coping With Fundus Images, Lan Chen, Yun Zhou, Songyang Gao, Mei Li, Hua Tan, Zhijiang Wan, Frontiers in Neuroscience, 2023.

🎖 Honors and Awards

Scholarships

• 2024.11 Baosteel Scholarship Special Prize (only 15 nationwide)
• 2024.09 / 2022.09 Nanchang University First Prize Scholarship
• 2023.12 / 2022.12 Huawei “Smart Base” Scholarship (ratio 0.3%), MoE–Huawei Future Star

Selected Competition Awards (20+ provincial-or-above, including 9 national-level)

• 2024.05 Best Poster Award, 6th Symposium on Computational Imaging Technology and Applications (First Author)
• 2024.10 National Silver Award, China International College Student Innovation Competition
• 2024.11 National Bronze Award, 14th “Challenge Cup” Entrepreneurship Competition (Team Captain)
• 2023.12 First Prize, 5th Global Campus AI Algorithm Elite Competition (Team Captain)
• 2023.11 First Prize, 14th Blue Bridge Cup Special Competition
• 2023.07 H Prize, Mathematical Contest in Modeling (MCM, top 25%)
• 2023.05 First Prize, 13th MathorCup Mathematical Modeling Challenge (Team Captain)

📖 Educations

• 2025.09 - 2027.06 (expected) M.Phil. in Artificial Intelligence, The Hong Kong University of Science and Technology (Guangzhou). Advisor: Prof. Hui Xiong (IEEE/ACM/AAAI/AAAS Fellow).
• 2021.09 - 2025.06 B.Eng. in Artificial Intelligence, Ji Luan Academy (AI Experimental Class, top 5% selection), Nanchang University (Double First-Class).

🔬 Research Experience

• 2022.10 - 2025.06 Research Assistant, LIVE Lab, Nanchang University. Investigated diffusion-model priors for computational imaging (digital holography and scattering-media imaging) to improve reconstruction quality at reduced hardware cost. Advisors: Prof. Qiegen Liu and Wenbo Wan.
• 2024.01 - 2024.03 Research Intern (Winter Camp), Institute for AI Industry Research (AIR), Tsinghua University. Built stable-diffusion-based weather-scene generation and editing pipelines to synthesize training data for autonomous driving. Supervisor: Yan Wang.
• 2025.03 - 2025.08 AI Algorithm Research Intern, Beijing Aishi Future Technology. Researched skeleton retargeting for diffusion-based human-animation generation.

📜 Patents

• H. Xiong, Y. Xia, S. Gao. “Intelligent Encapsulation and Controlled Release System for Big Scientific Data Based on Multi-Stage Workflow and Stream Computing.” (Accepted, 2026)
• W. Wan, S. Gao, B. Wang, W. Xu, Q. Liu. “A Dual-Channel Prior-Constrained Multi-Distance Lensless Digital Holographic Reconstruction Method.” ZL 2024107267996. (Student First Author, Granted)
• W. Wan, S. Gao, Z. Zhang, W. Xu, L. Zhang, Q. Liu. “A Sparse Digital Holographic Image Reconstruction Method.” ZL 202410464769.2. (Student First Author, Granted)

🛠 Skills

• Programming: Python, PyTorch, C/C++, Flink SQL, Next.js, FastAPI
• AI/ML: LLM fine-tuning, LLM-based agents, diffusion models, streaming data processing, computer vision
• Tools: Docker, Git, Linux, Slurm, vLLM, cloud deployment