Disentangling Visuo-Tactile Foresight: Oracle-Guided Interface Discovery for World Action Models
This work proposes Asymmetric Phase-Local Future Memory (AFM), in which visual memory reads future vision, each tactile memory jointly attends to its own tactile stream and phase-aligned future vision, and cross-tactile access is blocked, and compares it with Modality-Isolated Future Memory (IFM), which removes visual-...