arXiv new/cross; adaptive latent world model · P1 · 2026-07-01
AdaJEPA：它和通用视觉自监督的关系在于：AdaJEPA 用闭环自监督 transition 做 test-time adaptation，补上 frozen latent world model 的分布偏移问题
中高相关；详见方法、贡献和实验边界。
AdaJEPA: An Adaptive Latent World Model arXiv new/cross; adaptive latent world model 原文链接
编号2606.32026优先级P1类别arXiv new/cross; adaptive latent world model会议arXiv new/cross; adaptive latent world model方法AdaJEPA 用闭环自监督 transition 做 test-time adaptation，补上 frozen latent world model 的分布偏移问题来源arXiv / OpenReview
先说结论。它和通用视觉自监督的关系在于：AdaJEPA 用闭环自监督 transition 做 test-time adaptation，补上 frozen latent world model 的分布偏移问题。 中高相关；详见方法、贡献和实验边界。
Figureure 13 · : Visual Shifts: The model is trained on the default PushT data, but tFigure 13: Visual Shifts: The model is trained on the default PushT data, but the test-time observations have salt-and-pepper noise. AdaJEPA consistently decreases prediction loss.这张图来自论文 PDF 的结构化抽取。当前用于辅助理解 AdaJEPA 的方法或实验，请结合正文精读段落一起看。Figureure 2 · : Planning Success under Shape Shifts (top) and Visual Shifts (bottom)Figure 2: Planning Success under Shape Shifts (top) and Visual Shifts (bottom). The ★ denotes unseen shapes and configurations. AdaJEPA consistently improves planning success across all settings, using only a single adaptation step per MPC replanning step. We extend the maximum number of steps to 30 to show the increasing trend of planning success of AdaJEPA. Comparison between frozen and AdaJEPA planning trajectories is in Appendix C, showing AdaJEPA consistently decreases prediction loss and leads to better planning.这张图/表用于判断 AdaJEPA 的实验收益来自哪里。重点看替换、消融或跨模型设置下趋势是否一致，而不是只看单个最高分。
核心问题
它和通用视觉自监督的关系在于：AdaJEPA 用闭环自监督 transition 做 test-time adaptation，补上 frozen latent world model 的分布偏移问题。
方法拆解
AdaJEPA 用闭环自监督 transition 做 test-time adaptation，补上 frozen latent world model 的分布偏移问题
