EmbodiSkill: Skill-Aware Reflection for Self-Evolving Embodied Agents

ArXi:2605.10332v1 Announce Type: new Embodied agents can benefit from skills that guide object search, action execution, and state changes across diverse environments. Since embodied environments vary across layouts, object states, and other execution factors, these skills must self-evolve from trajectories generated during task execution. However, existing skill self-evolution methods are mainly developed in digital environments and often convert trajectories into coarse skill updates.