Yilai Liu
About
I am Yilai Liu, currently pursuing a Master of Philosophy degree at the University of Hong Kong (HKU). I conduct my research at NICE Lab under the supervision of Prof. Hongyang Du. Before that, I received my B.Eng. degree from Beijing University of Posts and Telecommunications (BUPT).
Research
My research focuses on temporal representation modeling, with an emphasis on contextual retrieval, compression and consistency modeling in long video and mobile data.
Feel free to contact me at 2369475677@qq.com for any potential discussion, collaboration and opportunity.
Education
2025 - Present
The University of Hong Kong
Master of Philosophy, Department of Electrical and Computer Engineering
2021 - 2025
Beijing University of Posts and Telecommunications
Bachelor of Engineering
Selected Works
View allVideo Understanding & Generation
PreCoG: Preview-Edit for Efficient Video Correction in Generation-Time
Propose PreCoG, an efficient video correction framework combining low-cost preview assessment with training-free editing that accounts for source-condition misalignment during generation.
TempCloze: Can Video-LLMs Identify the Missing Middle?
Introduce TempCloze, a video cloze benchmark for evaluating visual temporal reasoning beyond language shortcuts by identifying the missing middle from beginning and ending clips.
SlotMem: Character-Addressable Internal Memory for Narrative Long Video Generation
Propose SlotMem, a character-addressable internal memory framework for preserving long-range character consistency in narrative long-video generation.
Mobile Usage Behavior Modeling
MIDiff: Tackling Sparsity and Imbalance in Mobile Usage Generation via Multivariate-Imaging Diffusion
Propose MIDiff, a multivariate-imaging diffusion framework for generating realistic and diverse mobile usage traces under sparsity and usage imbalance.
U-MASK: User-adaptive Spatio-Temporal Masking for Personalized Mobile AI Applications
Propose U-MASK, a uniform model for multi-tasks generazabiltity through spatio-temporal masks for contexts-retrival under sparse observations.
Poster: Enhancing Mobile Traffic Data Generation through Spatio-temporal Correlation Imaging
Address the challenge of users' daily mobile usage modeling under usage sparsity by correlation imaging for data augmentation.
Competitions & Awards
ISPRS TC I Contest on Remote Sensing - Rank 2/43 2024
China International College Students' Innovation Competition - National Second Prize 2024
Chinese Mathematics Competitions for College Students - National Second Prize 2022