Concept来源依赖的噪声采样
source-dependent noise sampling
A training mechanism that dynamically adjusts noise levels or sampling weights according to the data source, so that rollout data from different sources remain balanced between behavior cloning and world modeling signals.
Timeline (last 90 days)
Oct 14
Source-dependent noise sampling dynamically adjusts the noise level or sampling weights according to the data source, so as to keep the behavior cloning signal and the world modeling signal balanced during training.
Unverified50%