学术前沿

PAD: Prediction with Action – Visual Policy Learning via Joint Denoising Process

提出预测-动作联合去噪框架,将图像预测与动作输出在同一扩散过程中学习。