← back to paper
arxiv: 2605.01896 · 2 revisions
Divide and Conquer: Decoupled Representation Alignment for Multimodal World Models