Qixiang Chen is a second-year Ph.D. student in the Department of Data Science and AI at Monash University, supervised by Prof. Jianfei Cai and Dr. Jingwen Ye. Before that, he received his Bachelor of Advanced Computing (Honours), majoring in Computer Vision and Machine Learning, from the Australian National University.

His research interests include multimodal learning, 3D spatial reasoning, vision-language models, and video understanding.

Education

• Ph.D. in Information Technology, Monash University, Aug 2025 – Present

• Bachelor of Advanced Computing (Honours), The Australian National University, Jul 2022 – Jul 2025

Publications

Seek-and-View Reasoning for Multi-View Spatial Understanding

Qixiang Chen, Cheng Zhang, Fucai Ke, Chi-Wing Fu, Jianfei Cai, Jingwen Ye

arXiv preprint, 2026

A Seek-and-View reasoning paradigm that seeks a question-relevant view to make the spatial evidence directly observable, realized by Vantage, a training-free and model-agnostic framework coupling VLM reasoning with a 3D foundation model.

Paper Project Page Code

OpenView: Empowering MLLMs with Out-of-view VQA

Qixiang Chen, Cheng Zhang, Chi-Wing Fu, Jingwen Ye, Jianfei Cai

The 40th Conference on Neural Information Processing Systems (NeurIPS 2026), Evaluations & Datasets Track

Out-of-view (OOV) VQA asks MLLMs to reason about objects, activities, and scenes beyond the visible frame of an image. We build a VQA synthesis pipeline to construct OpenView-Dataset, and a human-annotated OpenView-Bench for evaluation.

Paper Code Data

VAU-R1: Advancing Video Anomaly Understanding via Reinforcement Fine-Tuning

Liyun Zhu, Qixiang Chen, Xi Shen, Xiaodong Cun

arXiv preprint, 2025

A reinforcement fine-tuning framework that strengthens MLLM reasoning for video anomaly understanding, together with VAU-Bench, the first large-scale Chain-of-Thought benchmark for anomaly reasoning.

Paper Project Page Code Data

Motion meets Attention: Video Motion Prompts

Qixiang Chen, Lei Wang, Piotr Koniusz, Tom Gedeon

The 16th Asian Conference on Machine Learning (ACML 2024)
Long Presentation (5.67%)

A plug-and-play motion prompt layer that turns frame differencing maps into attention maps, so video models focus on the motion that matters for fine-grained action recognition.

Paper Project Page Code

Projects

PainterApp: Stroke Painting Algorithms with Shader Enhancements

Course Project, COMP4610 Computer Graphics, ANU, 2024

Stroke-based painting algorithms with shader enhancements that render photographs as realistic digital paintings, in both 2D and 3D.

Report Code

Vehicle Image Translation: Adapting Synthetic Styles to Real-World Scenarios

Course Project, COMP4660 Neural Networks, Deep Learning and Bio-inspired Computing, ANU, 2023

CycleGAN-based translation of synthetic Vehicle-X images into the style of the real-world VeRi dataset, improving domain adaptation for vehicle recognition.

Report Code