VLCache: Computing 2% Vision Tokens and Reusing 98% for Vision-Language Inference

Ähnliche Einträge

• ToFe: Lagged Token Freezing and Reusing for Efficient Vision Transformer Inference
  von: Zhang, Haoyue, et al.
  Veröffentlicht: (2025)
• An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models
  von: Chen, Liang, et al.
  Veröffentlicht: (2024)
• Think Twice, Act Once: Token-Aware Compression and Action Reuse for Efficient Inference in Vision-Language-Action Models
  von: Tan, Xudong, et al.
  Veröffentlicht: (2025)
• Language-Guided Token Compression with Reinforcement Learning in Large Vision-Language Models
  von: Cao, Sihan, et al.
  Veröffentlicht: (2026)
• TPC-ViT: Token Propagation Controller for Efficient Vision Transformer
  von: Zhu, Wentao
  Veröffentlicht: (2024)