Jerry Huang

PhD Student, Computer Science

Carnegie Mellon University

Hi! I am a PhD student in the Computer Science Department at Carnegie Mellon University, where I'm very lucky to be advised by Professor Nicholas Boffi.

My research is on generative modeling, where I have lately been focusing on diffusion and flow-based models. More broadly, I am interested in generative reasoning, representations, and multimodal intelligence, and building models that are efficient, controllable, and capable of reasoning across modalities such as images, video, language, and 3D.

Prior to CMU, I received my B.S. in Applied and Computational Mathematics from Caltech, where I was fortunate to be mentored by Professors John Preskill, Hsin-Yuan Huang, Adam Wierman, and Nicolas Christianson on topics in quantum computing and online algorithms.

Outside of research, I can be found playing tennis, watching my agents go brrrr, or attempting various new hobbies. Recently, I learned some photography, guitar, and dance.

Feel free to reach me at jerryhua [at] andrew [dot] cmu [dot] edu.

Projects

LTNT — a tool for exploring the latent space of diffusion models. Instead of one-shot prompting, you generate a spread of diverse interpretations of a prompt, then prune and breed toward the images you like — navigating the model's manifold rather than guessing prompts. A map view lets you fly through that space in 3D, organized by visual similarity. Built on GLASS interactive sampling over self-consistent flow-matching models (FLUX, SANA, Krea‑2).

Code Interactive app — coming soon (self-host with a GPU today)

Publications