I am an undergraduate student at the University of Michigan, Ann Arbor, pursuing a B.S.E. in Computer Science from the University of Michigan and a B.S.E. in Electrical and Computer Engineering from Shanghai Jiao Tong University (expected 2028). My research interests are multimodality, reasoning, representation learning, embodied AI, and brain-computer interfaces.
I am currently open to internship and research assistant opportunities.
我目前就读于密歇根大学安娜堡分校，攻读密歇根大学计算机科学工学学士与上海交通大学电子与计算机工程工学学士双学位（预计 2028 年毕业）。我的研究兴趣包括多模态、推理、表示学习、具身智能与脑机接口。
我目前开放实习与研究助理机会。
TacForcing: Streaming Action Generation with Execution-Time Tactile Feedback
arXiv preprint arXiv:2608.25798, 2026arXiv 预印本 arXiv:2608.25798，2026
Lets a robot react to what it feels while it moves, instead of only planning from what it felt beforehand. 让机器人在动作进行当中就能对触感做出反应，而不是只凭动作开始前摸到的东西来规划。
Reason, Ground, and Render: Consequence-Grounded Counterfactual Object Removal in Video
Under review, 2026审稿中，2026
Removes an object from a video and fixes what it leaves behind — the shadow it cast, the things it was holding up. 从视频里移除一个物体，并把它留下的痕迹一并补好——它投下的影子、它原本撑着的东西。
International Conference on Learning Representations (ICLR), 2026International Conference on Learning Representations (ICLR), 2026
Teaches a small vision-language model to read clinical EEG recordings and explain what it sees in plain words. 教一个小的视觉-语言模型看懂临床脑电图，并用平实的语言说清楚它看到了什么。
Annual Conference of the Cognitive Science Society (CogSci), 2026Annual Conference of the Cognitive Science Society (CogSci), 2026
Everyone's brain signals look different — this adapts an emotion model to a new person in seconds. 每个人的脑电信号都不一样——这个方法能在几秒内把情绪识别模型适配到新的人身上。
Honors Calculus II荣誉微积分 II
Academic Writing II学术写作 II
Introduction to Circuits电路导论