Yiqing Liang
Brown University
yiqing_liang@brown.edu
Bio
Yiqing Liang is a final-year PhD candidate at Brown University, advised by Prof. James Tompkin. Her research centers on spatiotemporal intelligence, enabling machines to sense, model, and reason about dynamic 3D/4D worlds through advances in dynamic scene reconstruction, geometric foundation models, and multimodal large language model reasoning. She earned her Master’s degree in Computer Science from Columbia University, where she worked with Prof. Shuran Song and Prof. Shih-Fu Chang, and completed her Bachelor’s degree in Computer Science at Fudan University in China. Yiqing is a recipient of CVPR 2025 Oral & Best Paper Candidate award. During her PhD, she has gained industry research experience through internships at NVIDIA Research and Meta Reality Labs.
Areas of Research
- Computer Graphics and Vision
Sense, Model, Reason: Toward Spatiotemporal Intelligence
My research aims to endow machines with spatiotemporal intelligence: the ability to perceive and reason about how the 3D world changes over time. I develop methods that close the loop between sensing, modeling, and reasoning. For sensing, ZeroMSF achieves zero-shot 3D motion estimation across diverse domains. For modeling, GauFRe advances dynamic Gaussian Splatting for single-camera 4D reconstruction, and MonoDyGauBench offers the first rigorous benchmark for dynamic NeRF/GS methods. For reasoning, SAFF integrates semantic features into dynamic radiance fields, while MoDoMoDo improves multimodal LLM reasoning via optimized data mixtures.
Going forward, I will couple 4D reconstruction with generative priors for prediction, infuse large transformers with physical awareness for interactive scene manipulation, and close the perception–action loop through active learning. These efforts will enable agents that perceive, predict, and act safely and efficiently in complex, real-world environments.