👋I’m Yuhan Wu (吴钰晗), a Ph.D. candidate in Computer Science at Peking University, advised by Prof. Tong Yang. I expect to graduate in 2026.
My research interests center on developing innovative data structures (数据结构) 📚 for networks, databases, and storage —an area where I’ve already achieved notable results. More recently, I’ve turned my focus to large language models training (大语言模型训练) 🚀, a field in which I’m actively making progress.
📧 Email: yuhan.wu AAT pku.edu.cn
💬 WeChat: yuhanwuuu
Education 🎓
| Degree | Institution | Major | Year |
|---|---|---|---|
| B.S. 本科 | Peking University 北京大学 | Electronics Engineering and Computer Science 信息科学技术学院 | 2017 – 2021 |
| Ph.D. 博士 | Peking University 北京大学 | Computer Science 计算机学院 | 2021 – 2026 (expected) |
Recent Practice (2024)
| English | 中文 |
|---|---|
| Large Model MOE Structure Optimization Cross-layer shared experts enable a 40% parameter reduction on an 8*0.5B model without compromising accuracy or activation. | 大模型MOE结构设计优化 跨层共享专家，在8*0.5B模型上节约40%参数量，同时保持精度和激活量不变。 |
| O1-like Model Model distillation, chain-of-thought construction, SFT, and enhancement of mathematical and coding abilities. | O1-like模型 模型蒸馏、思维链构建、SFT、数学代码能力提升。 |
| STEM Domain-Specific Large Model A large model designed to solve competition-level problems in STEM fields. | STEM垂直领域大模型 竞赛级难题解题。 |
| Tools for Large Models Training: Megatron (100× H800 GPUs); Finetuning: DeepSpeed, LLaMA-Factory; Evaluation: OpenCompass, vLLM. | 大模型相关工具使用 训练：Megatron（百卡H800）；微调：DeepSpeed、LLaMA-Factory；评测：OpenCompass、vLLM。 |
Fundings & Scholarships
| Fundings & Scholarships | 基金&奖学金 |
|---|---|
| NSFC Grant (300,000 RMB) (Top 1%) | 国家自然科学基金博士研究生项目，30万元，北大计算机Top5 |
| National Scholarship (China) (Top 1%) | 研究生国家奖学金，北大计算机Top 9 |
| Presidential Scholarship (Top 1%) | 北京大学“博士研究生校长奖学金”，北大计算机Top 4 |
| Huawei Scholarship | 华为奖学金 |
| Huawei Olympus Award | 华为奥林帕斯先锋奖 |
| Tun Scholarship | 北京大学董氏奖学金 |
| Wei Lin Scholarship | 北京大学巍璘奖学金 |
Awards
| Awards | 获奖 |
|---|---|
| Beijing Outstanding Undergraduate Thesis (Top 1%) | 北京市优秀本科毕业设计，全校35人入选，北大前1% |
| PKU Junior Scholar (Top 1%) | 北京大学未名学士，拔尖计划，全校50人入选 |
| Best Reviewer Award, KDD 2023 | KDD 杰出审稿人奖 |
| Top 10 Undergraduate Thesis Award, PKU EECS | 北京大学信息科学技术学院“十佳”本科毕业设计 |
| Outstanding Undergraduate Research Award, PKU | 北京大学本科生科研训练优秀项目奖 |
| Academic Innovation Award, PKU | 北京大学“学术创新奖” |
| Excellent Graduate, PKU | 北京大学优秀毕业生 |
| Pacemaker to Merit Student, PKU | 北京大学三好学生标兵 |
| Merit Student, PKU | 北京大学三好学生 |
Competitions
| Competitions | 竞赛 |
|---|---|
| NOI 2015 – Silver Medal | 全国信息学奥林匹克竞赛银牌 |
| NOI 2016 – Silver Medal | 全国信息学奥林匹克竞赛银牌 |
| ACM-ICPC Asia Regional Contest Shenyang Site 2018 – Silver Medal | ACM-ICPC 亚洲区域赛 沈阳站 2018 – 银牌 |
Publications
| No. | Authors & Title | Venue | Role |
|---|---|---|---|
| 1 | Yuhan Wu, Aomufei Yuan, Zhouran Shi, et al. “Online Detection of Outstanding Quantiles with QuantileFilter.” | ICDE 2024 | Independent First Author, CCF-A |
| 2 | Yuhan Wu, Feiyu Wang, Yifan Zhu, et al. “VisionEmbedder: Bit-Level-Compact Key-Value Storage with Constant Lookup, Rapid Updates, and Rare Failure.” | ICDE 2024 | Independent First Author, CCF-A |
| 3 | Yuhan Wu, Hanbo Wu, Chengjun Jia, et al. “Scalable Overspeed Item Detection in Streams.” | ICDE 2024 | First-ranked Co-First Author, CCF-A |
| 4 | Yuhan Wu, Shiqi Jiang, Siyuan Dong, et al. “MicroscopeSketch: Accurate Sliding Estimation Using Adaptive Zooming.” | KDD 2023 | First-ranked Co-First Author, CCF-A |
| 5 | Yuhan Wu, Jintao He, Shen Yan, et al. “Elastic Bloom Filter: Deletable and Expandable Filter using Elastic Fingerprints.” | IEEE Transactions on Computers 2021 | First-ranked Co-First Author, CCF-A |
| 6 | Yuhan Wu, Zirui Liu, Xiang Yu, et al. “MapEmbed: Perfect Hashing with High Load Factor and Fast Update.” | KDD 2021 | First-ranked Co-First Author, CCF-A |
| 7 | Yuhan Wu, Siyuan Dong, Yi Zhou, et al. “Kvsagg: Secure aggregation of distributed key-value sets.” | ICDE 2023 | First-ranked Co-First Author, CCF-A |
| 8 | Kaicheng Yang, Yuhan Wu, Ruijie Miao, et al. “ChameleMon: Shifting Measurement Attention as Network State Changes.” | SIGCOMM 2023 | Co-First Author, CCF-A |
| 9 | Jiarui Guo, Yisen Hong, Yuhan Wu, et al. “SketchPolymer: Estimate Per-item Tail Quantile Using One Sketch.” | KDD 2023 | Co-First Author, CCF-A |
| 10 | Zhuochen Fan, Zhoujing Hu, Yuhan Wu, et al. “PISketch: Finding Persistent and Infrequent Flows.” | IEEE/ACM Transactions on Networking | Co-First Author, CCF-A |