Ye Huang

Individual token selection and metadata caching unlock sparse attention for diffusion transformers without sacrificing generation quality, achieving up to $2.32\times$ speedups.

A novel decoupled autoencoder design unlocks 32x spatial compression for diffusion models while simultaneously boosting reconstruction fidelity and accelerating training convergence.

Runtime load balancing in FVAttn slashes attention processing time by over four times, transforming video generation efficiency.

Physically aligned video models can boost robotic manipulation success rates by over 50% compared to traditional methods.

Egocentric human video can outperform traditional teleoperated robot data, achieving superior performance in embodied model pretraining with lower costs and greater diversity.

Current Vision-Language Models get stuck in Pokemon Legends: Z-A not because they can't plan, but because they can't get unstuck.