Yuer Tang
"Build robots that learn. Understand intelligence along the way." — me
Hey there! I am a fourth-year undergraduate at UCLA studying Applied Mathematics and Linguistics & Computer Science, with a minor in Data Theory and Data Science Engineering. I'm broadly interested in how robots can learn and reason—a fascination that grew out of a deeper curiosity about the nature of intelligence itself.
I'm fortunate to receive mentorship from Prof. Yingnian Wu from the UCLA Statistics Department, Dr. Justin Baker from the UCLA Mathematics Department, and Dr. Jiayuan Mao from MIT.
Beyond research, I care about writing. It's been a while since I've written regularly, and I'm thinking about picking it back up.
News
January 2026: Paper accepted to the AAAI 2026 NeuroAI Workshop. See you in Singapore!
January 2026: Presented "When Linear Models Aren't Enough" at the Joint Mathematics Meeting in Washington D.C.
September 2025: Submitted work on coherent memory structures in neural fields to ICLR 2026.
May 2025: Started working with Jiayuan Mao at MIT CSAIL. My first time working on robotics!
Research
I'm interested in building robotic agents that learn and plan in ways that support people. My work spans representation learning for robot control, operator-theoretic methods for neural dynamics, and Bayesian approaches to goal inference.
Developing compact, interpretable latent representations for fine-grained robotic manipulation. Learning parameterizations where a small number of latent dimensions can continuously control policy scales (e.g., door-opening angle, motion speed) while preserving smoothness and interpretability.
Improving multi-task performance of Latent Plan Transformers by integrating Model-Agnostic Meta-Learning (MAML). Enabling agents to generalize across manipulation tasks while producing stable, continuous control through learned strategy latents.
Developing intrinsically explainable AI architectures using biologically inspired neural field representations. Framework leverages coherent structures for interpretable memory and decision-making.
Applied Mori-Zwanzig projection operator formalism to separate neural dynamics into resolved and unresolved states. Achieved major MSE reduction on neural data (awake state: 3.140 → 0.101) compared to baseline DMD.
Modeling and inferring agent goals within grid-based environments using Bayesian inference and reinforcement learning. Built full-stack online visualization system for real-time goal inference.
Developed a lightweight attention-enhanced backbone for 6-DoF camera pose estimation. Integrated CBAM with ResNet50, achieving 25% reduction in translation error and 35% reduction in rotation error.