Ziwei Liu

School of Data Science, City University of Hong Kong · Hong Kong, China · lziwei2-c@my.cityu.edu.hk

ziwliu8.github.io · GitHub · Google Scholar · ResearchGate

Ph.D. student in the AML Lab at the School of Data Science, City University of Hong Kong. My research focuses on Recommender Systems, Information Retrieval, and Large Language Models, with papers published or preprinted at top international AI conferences.

Ziwei Liu

Education

City University of Hong Kong

Ph.D. in Data Science. Affiliation: AML Lab

Sep 2025 - Present

City University of Hong Kong

Master of Engineering in Data Science. Supervisor: Prof. Zhao Xiangyu

Sep 2023 - Oct 2024

Southeast University

Bachelor of Engineering in Robotics Engineering. Co-supervised by Prof. Gan Yahui and Prof. Li Jun. Excellent Graduation Project

Sep 2019 - Jun 2023

Research Experience

Chinese University of Hong Kong (Shenzhen)

Research Assistant

May 2024 - Dec 2024

Publications & Preprints

Conditional Memory Enhanced Item Representation for Generative Recommendation (ComeIR)

Preprint, arXiv, May 2026.

  • Identifies bottlenecks in current decoupled two-stage generative recommendation frameworks and leverages engram memory for stronger GR input and decoding.

The Best of the Two Worlds: Harmonizing Semantic and Hash IDs for Sequential Recommendation (H2Rec)

Accepted by KDD'26 ADS Track, May 2026.

  • Harmonizes semantic IDs with hash IDs to address collaborative overwhelming in sequential recommendation, with deployment at Red Book (Xiao Hong Shu) and reported commercial gains.

LLM-EDT: Large Language Model Enhanced Cross-domain Sequential Recommendation with Dual-phase Training

Accepted by SIGIR'26, April 2026.

  • Proposes a dual-phase training paradigm and LLM-based item augmenter to address domain transition, imbalance, and rough profiling issues in cross-domain sequential recommendation.

SIGMA: Selected Gated Mamba for Sequential Recommendation

Accepted by AAAI'25, December 2024.

  • Studies the application of Mamba structures to sequential recommendation and introduces a selective gated module with a lightweight GRU branch.

Tutorials

Honors and Awards

Academic Service