MUSE: Resolving Manifold Misalignment in Visual Tokenization via Topological Orthogonality

Open paper
Benchmark Model Rank Results
text-to-image-generation-on-wiseMUSE-3B#5Overall: 0.65
text-to-image-generation-on-wiseMUSE-1B#8Overall: 0.58
image-generation-on-wiseMUSE-3B#24Overall: 0.65
image-generation-on-wiseMUSE-1B#30Overall: 0.58

Rank counts only results with a code link.