Hi, I am Qianyun Yang (杨茜云).
Welcome to my homepage! I am a master's student in the School of Software at Shandong University, under the supervision of Prof. Liqiang Nie and Dr. Zixu Li. My research interests span multimodal understanding and retrieval, heterogeneous vector search, evidence-driven agents, and AI security.
My research connects multimodal learning with reliable systems: I work on composed image retrieval and robust vision-language representations, hybrid and heterogeneous ANN search, evidence-grounded Agent evaluation through trajectory compression, evidence routing, and grader design, as well as AI security and artifact elimination. Across these topics, I focus on representation robustness, retrieval quality, system efficiency, and verifiable evaluation.
I also enjoy building systems beyond papers. My recent work includes the Huawei general vector-search collaboration, Alibaba agent-quality diagnostics, China Telecom data-infrastructure security, and multimodal retrieval research with collaborators from Shandong University.
欢迎来到我的主页！我目前是山东大学软件学院硕士研究生，导师为 Prof. Liqiang Nie，并由 Dr. Zixu Li 共同指导。我的研究兴趣主要聚焦于多模态理解与检索、异构数据向量检索、证据驱动 Agent 推理与人工智能安全。
我的研究将多模态学习与可信系统相结合，涉及组合图像检索与鲁棒视觉-语言表征、混合及异构 ANN 检索、通过轨迹压缩/证据路由/评分标准实现的证据驱动 Agent 评测，以及人工智能安全与生成伪影消除。我持续关注表征鲁棒性、检索质量、系统效率和可验证评测。
除学术论文外，我也关注真实系统落地，近期参与了华为通用向量检索合作、阿里巴巴 Agent 质量诊断、中国电信数据基础设施安全，以及多模态检索方向的系列研究。
I am affiliated with the Intelligent Media Research Center (iLearn). I believe open-source research makes multimodal learning more reproducible and collaborative. Our lab's projects and papers are open-source; please visit iLearn Lab and feel free to share your feedback.
我隶属于智能媒体研究中心（iLearn），并相信开源研究能够提升多模态学习的可复现性与协作性。实验室论文代码与相关项目持续开源，欢迎访问 iLearn Lab 并提出宝贵意见。
Representative projects connected to my research directions.
以下项目对应我的主要研究方向
📝 Selected Publications
STABLE: Efficient Hybrid Nearest Neighbor Search via Magnitude-Uniformity and Cardinality-Robustness
💼 Industry Experience
Alibaba · Taobao Tmall Group · China E-commerce Business Group
Algorithm Engineer · Multimodal · 2026.05 – 2026.09
Agent Harness-driven multimodal quality diagnosis and self-evolution.
- Designed trajectory compression, evidence routing, and Attribution/Critic/Counterfactual Grader rubrics for 300+ node Agent chains.
- Completed batch evidence verification with nearly 95% high-value case reduction; shortened end-to-end analysis from tens of minutes to the minute scale.
- Built a severity-aware SOP alignment and fine-tuning loop with SFT samples and graded rewards for recurring service-quality cases.
China Telecom · Data Infrastructure & Security Business Department
Algorithm Developer · 2025.09 – 2026.05
Data cleaning, fragment indexing, and intelligent matching for network-security assets.
- Built a rule/Agent dual-path cleaning system for non-compliant asset fields, category imbalance, and false alarms.
- Designed fragment-indexing and semantic/fine-grained matching algorithms for asset-vulnerability retrieval.
- Improved search logic, recall quality, coverage, and maintainability in a production-oriented pipeline.
📖 Education
- 2024.09 – present · Master's student, School of Software, Shandong University.
- 2020.09 – 2024.07 · Bachelor's degree, School of Software, Shandong University.