ReFIT:指令引导的视觉 token 削减框架
Get the story
2026年10月8日,arXiv Computer Vision 频道发布论文,提出 ReFIT,一种面向高效大型视觉语言模型(LVLM)推理的指令引导视觉 token 削减框架,旨在加速图像理解。该框架由两个组件构成:Relevance-Guided Window Reshaping(RWR,相关性引导的窗口重塑)与 Instruction-Guided Token Refinement(ITR,指令引导的 token 精炼)。目前公开信息仅包含框架名称与组成,未披露具体实验数据、基准结果或代码开源情况。
Generated from reports · updated 3 hr ago
Timeline
Follow the coverage from different angles.
- arXiv · Computer VisionReFIT:面向加速图像理解的自适应视觉 token 削减
论文提出 ReFIT,一种面向高效 LVLM 推理的指令引导视觉 token 削减框架,由 Relevance-Guided Window Reshaping(RWR)与 Instruction-Guided Token Refinement(ITR)组成。
Heat trend
Current heat 9·Comparable peak 10(Oct 8)·Comparable change over 24 hours –
The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.