AI 精选动态
智能评分 60
Yann LeCun 转发 AdaJEPA 论文
AI 推荐理由
与已有自适应方法相比,AdaJEPA 结合模型预测控制,实现在测试时通过预测与观测匹配进行自适应,无需元学习。核心解读
Yann LeCun 及合作者提出 AdaJEPA 模型,实现测试时自适应,将 LeWorld 模型与模型预测控制结合,通过执行动作后对比预测的潜在状态与观测状态进行自适应。论文已发布。
全文
Yann LeCun (@ylecun) 转发了 Günter Klambauer (@gklambauer) 的帖子:
LeWorld model becomes ADAPTIVE and meets MODEL-PREDICTIVE CONTROL
AdaJEPA by Yann LeCun and colleagues performs actions, then checks the predicted latent state versus the observed and adapts at TEST TIME.
Similar to meta-learning ...
Paper: https://arxiv.org/abs/2606.32026 https://t.co/qaJZ9KvJy5
