Can We Achieve Efficient Diffusion without Self-Attention? Distilling Self-Attention into Convolutions

Documenti analoghi

• Delving into Cascaded Instability: A Lipschitz Continuity View on Image Restoration and Object Detection Synergy
  di: Zhao, Qing, et al.
  Pubblicazione: (2025)
• Language Generation as Optimal Control: Closed-Loop Diffusion in Latent Control Space
  di: Dong, ZiYi, et al.
  Pubblicazione: (2026)
• When Preference Labels Fall Short: Aligning Diffusion Models from Real Data
  di: Chen, Weiyan, et al.
  Pubblicazione: (2026)
• Slot Attention with Re-Initialization and Self-Distillation
  di: Zhao, Rongzhen, et al.
  Pubblicazione: (2025)
• Attentive Convolution: Unifying the Expressivity of Self-Attention with Convolutional Efficiency
  di: Yu, Hao, et al.
  Pubblicazione: (2025)