AI can generate stunning video, yet fails at stable physics and viewpoints. A new framework argues real world models require three kinds of consistency.
AI can generate stunning video, yet fails at stable physics and viewpoints. A new framework argues real world models require three kinds of consistency.