Local Information Matters: Inference Acceleration For Grounded Conversation Generation Models Through Adaptive Local-Aware Token Pruning

Documenti analoghi

• Think Twice, Act Once: Token-Aware Compression and Action Reuse for Efficient Inference in Vision-Language-Action Models
  di: Tan, Xudong, et al.
  Pubblicazione: (2025)
• MADTP: Multimodal Alignment-Guided Dynamic Token Pruning for Accelerating Vision-Language Transformer
  di: Cao, Jianjian, et al.
  Pubblicazione: (2024)
• Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization
  di: Li, Kaiyuan, et al.
  Pubblicazione: (2025)
• TokenCarve: Information-Preserving Visual Token Compression in Multimodal Large Language Models
  di: Tan, Xudong, et al.
  Pubblicazione: (2025)
• AdaptInfer: Adaptive Token Pruning for Vision-Language Model Inference with Dynamical Text Guidance
  di: Zhang, Weichen, et al.
  Pubblicazione: (2025)