Skip to content
Hot eventLive

HGSS:文本到图像扩散模型的组合概念擦除框架

1 reports1 sources3 hr ago updated

Get the story

AI overview

2026年10月7日,arXiv(Computer Vision)发布论文,提出面向文本到图像扩散模型的免训练组合概念擦除框架HGSS。该方法通过层次跨度定位与动态属性绑定来消除目标概念,特点是无需更新模型权重、不增加学习参数。这是目前该事件的唯一报道,尚未见后续验证、复现或争议信息。

Generated from reports · updated 3 hr ago

Timeline

Follow the coverage from different angles.

Oct 7, 2026
  1. arXiv · Computer Vision
    HGSS:文本到图像扩散模型中基于层次语义手术的组合概念擦除

    论文提出免训练框架 HGSS,用于文本到图像扩散模型的组合概念擦除,通过层次跨度定位与动态属性绑定消除目标概念且不更新模型权重、不增加学习参数。

Heat trend

Current heat 9·Comparable peak 10(Oct 7)·Comparable change over 24 hours –

02.557.510Oct7Oct7Oct7Oct7

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.