Yazhe Wan (万亚哲)

I'm currently a PhD student in the Centre for Intelligent Sensing within the School of Electronic Engineering and Computer Science at Queen Mary, University of London, supervised by Dr. Changjae Oh and Dr. Ziquan Liu. I joined this Ph.D. program after completing my undergraduate studies through the Joint Programme between Queen Mary University of London and Beijing University of Posts and Telecommunications (BUPT).

Email  /  Google Scholar  /  GitHub

Yazhe Wan

Current Research (Ph.D.)

My research interests lie in embodied AI, robotics, and computer vision. I am currently working on multi-modal foundation models for generalizable robotic manipulation.

Publications

Point-wise vs. relational alignment between 3D and 2D feature spaces
Relational Spatial Alignment for Vision-Language-Action Models
Chaoran Zhu, Hengyi Wang, Chenglin Cui, Yazhe Wan, Taein Kwon, Changjae Oh
Conference on Robot Learning (CoRL), 2026

Aligning relational structure, rather than individual features, between 3D and 2D feature spaces for spatially-aware VLA models.

Open-vocabulary detection results of DAT
The Detector Teaches Itself: Lightweight Self-Supervised Adaptation for Open-Vocabulary Object Detection
Yazhe Wan, Changjae Oh
International Conference on Pattern Recognition (ICPR), 2026
paper / arXiv / code / bibtex

A plug-and-play, self-supervised fine-tuning of the VLM backbone (<0.8M parameters) that improves region-level features for cooperative open-vocabulary detection.

EVA-Score: extraction, validation, information score
EVA-Score: Evaluating Abstractive Long-form Summarization on Informativeness through Extraction and Validation
Yuchen Fan, Yazhe Wan, Xin Zhong, Haonan Cheng, Ning Ding, Bowen Zhou
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026
paper / arXiv / bibtex

An information-based metric for long-form summaries that extracts the information in a summary and validates it against the reference.