Zhengming Yu

Adversarial distillation achieves state-of-the-art few-step visual generation by learning log-density ratios directly via classification, eliminating the need for auxiliary score fitting.

Breathe new life into dull, old videos: DiffHDR uses video diffusion models to hallucinate realistic HDR detail from LDR footage, unlocking unprecedented re-exposure latitude.

Achieve both fidelity and plausibility in 3D reconstruction by unifying feed-forward reconstruction with diffusion-based generation in a single cooperative system.