I'm a Research Scientist at Meta in Menlo Park. I think about how to generate things. My research is on building simplified abstractions of the world through the lens of dynamical systems and flows.
Lately, I've been exploring insertion-based sequence generation [Edit Flows], investigating scaling laws and new capabilities for multimodal generation [OneFlow]. Among my research outputs, data-driven methods such as [Flow Matching] have been applied successfully by many for foundation models of video and audio [Movie Gen, SD3, etc]. Additionally, reward-driven methods such as [Adjoint Matching] have been applied to large-scale diffusion finetuning for internal GenAI models and AI & Chemistry [AS/ASBS].
CV | Github | Twitter | Google Scholar | rtqichen@gmail.com