Quantitative Researcher · Ubiquant
I’m a quantitative researcher at Ubiquant. I joined in July 2026, following several years of research in computer vision and multimodal learning.
Previously, I worked on multimodal models at Ant Group and autonomous 3D characters with MMLab at NTU and SenseTime. This page brings together some of that work.
Real-User Behavior-grounded Rubric for Multimodal Interaction Capability Evaluation
We study the gap between benchmark scores and real conversations, evaluating both individual responses and whole sessions, with analysis to help explain where models fall short.
Camera-ready in preparation; paper and materials will follow.
Ming-flash-omni Preview · Ming-flash-omni 2.0
At Ant Group, I worked on video understanding and audiovisual interaction, including training data, post-training, and evaluation.
Video understanding, audiovisual interaction, data synthesis, and automated evaluation.
Working at Ant Group was a truly terrible experience for me.
Autonomous 3D characters, social intelligence, and multimodal interaction.
Camera Intelligence Lab · Event-based perception and hand interaction. Research collaboration with Prof. Boxin Shi.
2023 · BAAIResearch intern · Working with Yue Cao and Xinlong Wang.
2021 · ByteDanceAI product management intern · Intelligent video creation and AR research.
2019–2021 · AREYECore founding-team member · AR experiences for museums and cultural spaces.
Reflections on research, personal growth, and life at the end of graduate school.
An early look at the limits of AR hardware, from optics and displays to interaction.
WeChat public account
You can also find my writing on WeChat. Scan the code to follow.
A reading list on language models in financial research and trading, collected as I explored the field.
For research discussions or to get in touch, you can reach me by email.