Custom Forcing:免训练自回归视频主体定制
Get the story
2026年10月5日,arXiv(Computer Vision,一手来源)报道 Custom Forcing,一种免训练的自回归视频生成主体定制方法。其核心思路是把参考锚帧存入冻结的 autoregressive 视频模型的持久 KV cache,从而实现主体定制。方法包含两个机制:一是 drift-adaptive value amplification(DVA),随身份漂移程度放大参考影响;二是 anchor contrast guidance(ACG),引导生成偏离通用类别先验。目前公开信息仅涉及该方法的设计思路与组件,未见实验数据、代码或第三方验证。
Generated from reports · updated 1 days ago
Timeline
Follow the coverage from different angles.
- arXiv · Computer VisionCustom Forcing:免训练的 autoregressive 视频生成主体定制方法
Custom Forcing 是一种免训练方法,把参考锚帧存入冻结的 autoregressive 视频模型的持久 KV cache,实现主体定制。它用 drift-adaptive value amplification(DVA)随身份漂移程度放大参考影响,并用 anchor contrast guidance(ACG)引导生成偏离通用类别先验。
Heat trend
Current heat 5·Comparable peak 10(Oct 5)·Comparable change over 24 hours -50%
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.