3

Perceptio explores perception-enhanced vision-language modeling through spatial token generation for complex 2D and 3D spatial reasoning.

This work studies music-driven dance video generation with long-horizon pose diffusion, focusing on robust motion synthesis and pose alignment for video generation.