NYU
academicNew York University's AI research centers on the Courant Institute and the Center for Data Science (CDS). The most consequential group of the window is Saining Xie's vision-LLM lab, whose Cambrian-1 vision-centric multimodal family and its successors (Cambrian-S, Cambrian-P) drive the spatial-supersensing agenda alongside the VSI-Bench video spatial-intelligence benchmark. The same group produced the diffusion-representation line — REPA (ICLR 2025 oral), RAE, and Scale-RAE — that reshaped how diffusion transformers use pretrained visual encoders.
Breadth is a hallmark: Solaris is the first multiplayer video world model; Eugene Vinitsky's Emerge Lab ships the 1M-FPS GPUDrive multi-agent driving simulator; and the Pinto/Shafiullah robotics group's Robot Utility Models are zero-shot-deployable general robot policies. Yann LeCun remains an NYU professor and Kyunghyun Cho and He He anchor the NLP/ML foundations side.
Industry mobility is intense at the top: LeCun left Meta in November 2025 and co-founded AMI Labs (announced 2026-03, $1.03B seed), with Xie as co-founder and Chief Science Officer — AMI Labs output is tracked as a separate lab, not NYU. Cho holds a joint role at Prescient Design (Genentech), where his bio-FM work ships. The entry keeps only NYU-branded artifacts, so the JEPA line, MetaMorph, and Web-SSL (Meta-branded) are excluded despite NYU co-authors.
People
- Yann LeCun Google Scholar — Silver Professor, Courant/CDS; Turing Award laureate (formerly Chief AI Scientist, Meta (until 2025); Co-Founder, AMI Labs (2026))
- Kyunghyun Cho Google Scholar — Professor, Courant/CDS (formerly Senior Director of Frontier Research, Prescient Design (Genentech))
- Saining Xie Google Scholar — Associate Professor, Courant; led Cambrian/REPA/RAE/VSI line (formerly Co-Founder & Chief Science Officer, AMI Labs (2026); prior Google DeepMind)
- He He Google Scholar — Associate Professor, Courant/CDS; LLM reasoning, truthfulness, safety
- Eugene Vinitsky Google Scholar — Assistant Professor, NYU Tandon; Emerge Lab (multi-agent RL)