Runqian (Ray) Wang

I am an undergraduate student at MIT double majoring in AI and Math. I am currently advised by Kaiming He at MIT and Yilun Du at Harvard. Previously, I was advised by Aude Oliva, Polina Golland, Zhirong Wu, and Rogerio Feris. I also interned as a student researcher at Microsoft Research and IBM Research.

Email  /  CV  /  Github  /  Google Scholar

My current focus is on generative modeling. I used to work on computer vision, optimization, LLM finetuning, and medical imaging. I am most interested in simple yet effective methods that can generalize and scale.

New generative model with equilibrium dynamics that exceeds Flow Matching in performance, supports optimization-based sampling, and naturally performs multiple downstream tasks.

Plug-and-play representation regularizer for generative modeling that brings consistent improvement across different settings.

Enables nearly data-free and compute-efficient transfer of existing PEFT modules trained on old base model to new base models, while at least preserving, in most cases improve, performance.

Provides a comprehensive study on malposition detection using deep learning approaches and proposes a new approach for malposition classification using Mask-RCNN and SWIN transformer.

Proposes a new algorithm to compute the X-ray transform of an image; improves time complexity from O(N^d) to O(N).

Applies Grover's quantum search algorithm to solve satisfiability problems and reduces time complexity to the square root of existing solutions.