Ziqi Ma

Computer Vision @ California Institute of Technology.

headshot1.JPG

Hello! I am Ziqi, a PhD student at Caltech. My goal is to let AI understand and act in the physical world. Currently I do this via “world modeling’’ – letting AI learn physical priors by generating and modeling the world. I am advised by Georgia Gkioxari and Yisong Yue. I interned at World Labs and Meta SuperIntelligence Labs. I am also a Microsoft and UChicago alumni.

News

Sep 24, 2026 DynaTokens got accepted to NeurIPS. See you in Atlanta!
Sep 01, 2026 Atlas is released! Best camera controlled video gen, plus everything you get with an omni AR model!
Jun 17, 2026 Out of Sight, Out of Mind? Evaluating State Evolution in Video World Models is accepted to ECCV!
Jun 08, 2026 Started my internship at World Labs – my first startup experience!
Jun 06, 2026 SAM 3D got CVPR 2026 Best Paper Award Honorable Mention!

Selected Publications

  1. DynaTokens: Teaching Dynamics to Camera-Controlled Video Models at Test Time
    Ziqi Ma, Hongqiao Chen, and Georgia Gkioxari
    NeurIPS, 2026
    PDF Website
  2. Out of Sight, Out of Mind? Evaluating State Evolution in Video World Models
    Ziqi Ma*, Mengzhan Liufu*, and Georgia Gkioxari
    ECCV, 2026
    PDF Blog Website
  3. sam3d.png
    SAM 3D: 3Dfy Anything in Images
    SAM 3D Team
    CVPR · Best Paper Award Honorable Mention ✨, 2025
    PDF Blog Website
  4. steer3d.png
    Feedforward 3D Editing via Text-Steerable Image-to-3D
    Ziqi Ma, Hongqiao Chen, Yisong Yue, and Georgia Gkioxari
    2025
    PDF Website
  5. find3d.png
    Find Any Part in 3D
    Ziqi Ma, Yisong Yue, and Georgia Gkioxari
    ICCV · Highlight ✨, 2025
    PDF Blog Website
  6. uq_teaser.png
    Calibrated Uncertainty Quantification for Operator Learning via Conformal Prediction
    Ziqi Ma, David Pitt, Kamyar Azizzadenesheli, and Anima Anandkumar
    TMLR, 2024
    PDF

Latest Posts

Mar 13, 2026 Out of Sight, Out of Mind: Can Video World Models Simulate A World Beyond the Pixel Frame?
Dec 28, 2025 Rodney Brooks, Tolstoy, and World Models
Aug 06, 2025 Find3D: Localizing Semantic Concepts in the 3D Space