From Video Generation to World Model

Introduction

In recent years, the research community has made significant strides in generative models, particularly in the area of video generation. Despite challenges in generating temporally coherent and physically realistic videos, recent breakthroughs such as SORA, Genie, and MovieGen show promising progress toward controllable, high-fidelity visual world models. This tutorial offers a deep dive into recent advances in text-to-video generation, diffusion-based video models, and the bridge from generative video to physical and interactive world modeling. We aim to provide attendees with a comprehensive understanding of these cutting-edge methods and how they contribute to building embodied world models.

Schedule

Time (GMT-5)	Programme
09:20 - 09:30	Opening Remarks
09:40 - 10:20	Invited Talk: The Placeholder of the Talk Title Jack Parker-Holder Research Scientist, Google DeepMind
10:20 - 10:40	Coffee Break
10:40 - 11:20	Invited Talk: The Placeholder of the Talk Title Hong-Xing "Koven" Yu Ph.D. candidate at Stanford University
11:20 - 13:30	Lunch Break
13:30 - 14:10	Invited Talk: Breaking the Algorithmic Ceiling in Pre-Training with an Inference-first Perspective Jiaming Song Chief Scientist at Luma AI
14:10 - 14:20	Coffee Break
14:20 - 15:10	Invited Talk: The Placeholder of the Talk Title Pengfei Wan Head of KLing AI, Kuaishou
15:20 - 15:30	Coffee Break
15:30 - 16:00	Invited Talk: The Placeholder of the Talk Title Angjoo Kanazawa Assistant Professor, UC Berkeley
16:00 - 16:10	Coffee Break
16:10 - 16:50	Generative World Modeling for Embodied Learning Sherry Yang Assistant Professor, New York University
16:50 - 17:00	Ending Remarks (Lucky Draw)