Taichi Liu

I am a Computer Science PhD student at Rutgers University, working on Spatial Intelligence, Embodied AI and Machine Learning, under the supervision of Prof. Desheng Zhang.

Previously, I was a Master's student in Biomedical Engineering at Johns Hopkins University and in Electrical Engineering at Tsinghua University. My research focused on Machine Learning and Recommender Systems, with an emphasis on Cold-start Recommendation, Consumption Intention Modeling, and Conversational Personalized Recommendation.

I was also interested in topics from Personalized Machine Learning and in 2023 I worked as a Research Assistant at UCSD, at Julian McAuley's Lab.

Email / CV / Google Scholar / Linkedin

News

September 2025: OP3Det is accepted at NeurIPS 2025 in San Diego!
July 2024: Our work BHIPP was accepted at CIKM 2024 in Idaho!
July 2024: Our work Ov-uni3detr was accepted at ECCV 2024 in Milano!
September 2023: I started my PhD in Computer Science at Rutgers University!
April 2023: UCC was accepted at SIGIR 2023 in Taipei!
May 2022: User Consumption Intention Prediction Model from my internship at Meituan was accepted at KDD 2022!

Selected Research

	Towards 3D Objectness Learning in an Open World Taichi Liu, Zhenyu Wang, Ruofeng Liu, Guang Wang, Desheng Zhang NeurIPS* 2025 project page / paper OP3Det is a class-agnostic open-world prompt-free 3D detector to detect any objects within 3D scenes.
	Behavior-Aware Hypergraph Convolutional Network for Illegal Parking Prediction with Multi-Source Contextual Information Zhenyu Wang, Yali Li, Taichi Liu, Hengshuang Zhao, Shengjin Wang CIKM 2024 paper BHIPP is a novel behavior-aware hypergraph convolutional network for city-wide illegal parking prediction.
	OV-Uni3DETR: Towards Unified Open-Vocabulary 3D Object Detection via Cycle-Modality Propagation Zhenyu Wang, Yali Li, Taichi Liu, Hengshuang Zhao, Shengjin Wang ECCV 2024 arXiv / code OV-Uni3DETR is a unified multi-modal open-vocabulary 3D detector that recognizes and localizes novel classes well, achieving modality unifying and scene unifying.
	Uncertainty-aware Consistency Learning for Cold-Start Item Recommendation Taichi Liu, Chen Gao, Zhenyu Wang, Dong Li, Jianye Hao, Depeng Jin, Yong Li SIGIR 2023 paper UCC allows the cold items to have similar distribution with the warm items by additionally generating low-uncertainty interactions.
	User Consumption Intention Prediction in Meituan Yukun Ping, Chen Gao, Taichi Liu, Xiaoyi Du, Hengliang Luo, Depeng Jin, Yong Li KDD 2022 paper GRIP is a graph neural network-based intention prediction model which can capture user intrinsic preference and spatio-temporal context.