Search papers, labs, and topics across Lattice.
Shanghai Jiao Tong University
4
0
5
Transforming continuous latent states into discrete tokens enables up to 20脳 compression while enhancing interpretability in reasoning tasks.
ICRDrag achieves superior editing accuracy and visual fidelity by redefining region-based drag in diffusion models, leaving point-based methods in the dust.
PermaVid achieves unprecedented long-term consistency in video generation, even after significant edits, by intelligently disentangling appearance and geometry in memory.
Achieving an overall score of 84.76, DreamX-World 1.0 sets a new benchmark for interactive video generation, outperforming leading models in both camera control and visual fidelity.
