Model
AtomicVLA
A robotic learning approach that combines skill discovery with a VLA framework, achieving skill grouping through action sequence clustering. It relies primarily on surface-level similarity rather than behavioral semantic equivalence.
Timeline (last 90 days)
Sep 11
Methods such as AtomicVLA and AtomSkill attempt to group and categorize action sequences, but rely on surface-level similarity or heuristic rules, sidestepping the essential judgment of behavioral equivalence.
Unverified50%