Achievement | Associate Professor Pengkun Yang Receives Huawei Spark Award

Release Time:2026-08-03 16:53:20

Huawei Technologies recently announced the list of recipients of the 144th round of its "Spark Award" for solving challenging problems. Associate Professor Pengkun Yang from the Department of Statistics and Data Science at Tsinghua University was awarded the Huawei "Spark Award" for the high value of a spatiotemporal sequence foundation model framework for networks with an In-Context Learning paradigm that he proposed in tackling the Zhufeng challenge problem "Spatiotemporal Sequence Foundation Model Technology for Networks."

Professor Pengkun Yang led a team comprising Songtai He (supervisor: Pengkun Yang; Ph.D. student, Class of 2026, Department of Statistics; research direction: statistical machine learning) and Yijiong Zhang (postdoctoral researcher; collaborating supervisor: Dong Li; research direction: time series) in a joint research endeavor. Targeting key technical challenges in the field of network operations—including the strong heterogeneity of massive KPI spatiotemporal sequences, the difficulty of unified representation, and the high cost of multi-task modeling—the team proposed a structure-aware zero-shot in-context inference paradigm for spatiotemporal sequences. The approach constructs learnable, interpretable, and transferable spatiotemporal representations based on statistical, periodic, and state features, and combines them with a prior data fitting model to achieve unified inference across tasks such as single-KPI forecasting, anomaly detection, and multi-KPI root cause localization, providing an innovative general-purpose technical solution for intelligent network operations.


Profile of Associate Professor Pengkun Yang

Pengkun Yang serves as a "Minxing Scholar" Associate Professor in the Department of Statistics and Data Science at Tsinghua University. His long-standing research focuses on high-dimensional statistics and the theory and algorithms of statistical learning. He has produced a series of innovative achievements addressing issues including statistical modeling of complex heterogeneous data, optimal statistical inference, and latent structure learning.


About Huawei's "Spark Award"

The Spark Award derives its name from the notion of sparks striking against one another—the spark of colliding ideas. Huawei places a premium on basic research and adheres to a philosophy of open innovation. Since November 2021, the company has distilled the persistent, deeply challenging problems arising from its industrial campaigns and from across the sector into scientific questions of both commercial and scientific value, and made them publicly available. Through a challenge-solicitation mechanism, Huawei seeks individuals with rigorous theoretical grounding and innovative problem-solving approaches. In recognition of, and gratitude for, the substantial contributions that challenge solvers make to both industry and the scientific community, Huawei established the "Spark Award."

Edited by Yushen Hou, Songtai He

Reviewed by Lin Hou, Lefei Li