Apple AI Updates: August 1, 2026

1. Apple’s MoMo Learns Motion Modes as Reusable Robot Manipulation Behaviors

Apple. Apple ML researchers introduced MoMo, a two-stage imitation-learning framework that treats execution-level motion variation as a reusable behavioral factor across robot manipulation tasks. It pairs a spatiotemporal action tokenizer that encodes motion patterns with a behavior-cloning transformer conditioned on task and continuous motion-mode inputs, producing distinct steady, dynamic, and intermediate behaviors. The system showed compositional generalization, transferring unseen motion modes to tasks trained in a single mode while maintaining success rates. Source