My name is Shaoyu Liu, and I am currently a Ph.D. candidate jointly trained by Xidian University and the Qiyuan National Laboratory, with an expected graduation year of 2027. My advisors are Professor Guanghui Zhao (Xidian University) and Professor Xiangyang Ji (Qiyuan Lab, Tsinghua University). To date, I have published one first-author paper at CVPR2025 and ICLR2026 and have another papers under review at CVPR2026. During my studies, I was awarded the Qiyuan National Laboratory Academic Scholarship. I developed the first event-based vision multimodal large language model, EventGPT, which has garnered widespread attention both domestically and internationally (including recognition from the Robotics and Perception Group led by Davide Scaramuzza at the University of Zurich, as well as Arizona State University, among others). It was featured in a dedicated article titled “EventGPT: Giving AI the Power of Superhuman Vision” by the U.S. developer platform PromptLayer.

My primary research interests include multimodal large language models, neuromorphic vision, robotics and perception, and related fields.

🔥 News

• 2026.01:  🎉🎉 One Paper accepted by [ICLR2026].
• 2025.11:   EventBench Project Page is Now Live!.
• 2025.11:   Submitted a paper to [CVPR2026].
• 2025.09:   Submitted a paper to [ICLR2026].
• 2025.05:   Submitted a paper to [NeurIPS2025].
• 2025.02:  🎉🎉 One Paper accepted by [CVPR2025].
• 2024.11:   EventGPT Project Page is Now Live!.
• 2024.11:   Submitted a paper to [CVPR2025].

📝 Publications

CVPR 2025

EventGPT: Event Stream Understanding with Multimodal Large Language Models [CVPR2025]

Shaoyu Liu, Jianing Li, Guanghui Zhao, Yunjian Zhang, XinMeng, Xiangyang Ji, Ming Li

Project Page

• The first LLM tailored for event streams, enabling precise temporal reasoning and natural language understanding from asynchronous visual signals.

ICLR2026

EventFlash: Towards Efficient MLLMs For Event-Based Vision [ICLR2026]

Shaoyu Liu, Jianing Li, Guanghui Zhao, Yunjian Zhang, Wen Jiang, Ming Li, Xiangyang Ji

• The efficient LLM tailored for event streams, enabling efficient temporal reasoning and dynamic scene understanding from asynchronous event stream.

ArXiV

EventBench: Towards Comprehensive Benchmarking of Event-based MLLMs [ArXiv] Submit to CVPR2026

Shaoyu Liu, Jianing Li, Guanghui Zhao, Yunjian Zhang, Xiangyang Ji

Project Page

• The comprehensive benchmark for event streams, enabling structured evaluation of understanding, recognition, and spatial reasoning.

🎖 Honors and Awards

• 2025.09 🎉🎉 Received the Qiyuan National Lab Academic Scholarship
• 2024.09 🎉🎉 Received the Qiyuan National Lab Academic Scholarship

💬 Academic Services

Conference Reviewer

• IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
• European Conference on Computer Vision (ECCV)
• International Conference on Learning Representations (ICLR)

💻 Internships

• 2024.09 - 2025.03, Research Intern at GuangMing Lab, Shenzhen, China.
• 2024.03 - 2024.08, Algorithm Intern at Zhizi Xiyuan, Shanghai, China.