Yi Yang

Predictive pretraining for parametric PDEs unlocks a 51.4% generalization improvement on unseen governing parameters by explicitly modeling latent dynamics geometry.

Multi-step flow-matching policies can achieve single-step control latencies by pipelining partially denoised action trajectories through an asynchronous rolling buffer.

Bypassing RGB entirely, Latent-to-4D achieves significant improvements in 4D scene generation while maintaining a reusable framework across different video models.