CarryOn: A Dual-Loop Harness for Personalized and Reliable Robot Assistance
CarryOn personalizes frozen robot VLA policies and cuts hazardous violations while improving failure recovery.
CarryOn wraps a frozen vision-language-action policy with a planning loop that maintains user and scene context for personalized subtasks, and an execution loop that shields actions from predicted hazardous contact. When execution fails, reusable control-based recovery skills restore progress and are refined from earlier failures. On a real robot and in simulation, violations of an already safety-finetuned VLA fell from 52.6% to 6.1%, and recovery success rose from 13% to 80% when refined skills were reused.
- Planning loop keeps persistent user and scene context.
- Execution loop blocks actions predicted to cause hazardous contact.
- Safety-finetuned VLA violation rate fell from 52.6% to 6.1%.
- Recovery success rose from 13% to 80% with reused skills.
Full article192 words · extracted from huggingface.co · click to collapse
While generalist vision-language-action (VLA) models have enabled diverse open-world manipulation, real-world robotic assistance requires two critical capabilities they currently lack: personalization and reliable execution. First, executing an instruction like "make me a sandwich" requires grounding subtask decomposition in persistent user preferences rather than relying on a single generic solution. Second, reliable deployment demands explainable safety assurance that proactively prevents hazards and gracefully recovers from occasional physical failures. To address these gaps, we present CarryOn, a dual-loop harness designed to augment a frozen VLA policy. The planning loop maintains persistent user-level and scene-level context that is continually grounded against new observations and revised through interaction, enabling personalized task selection and execution grounding across conversations and changing scenes. The execution loop actively supervises policy rollouts, shielding actions from predicted hazardous contact. When execution fails, the system invokes reusable control-based recovery skills that restore task progress and are refined through reflection on earlier failures. On a real robot and in simulation, CarryOn dispatches preference-consistent subtasks across users and scenes, reduces the violation rate of an already safety-finetuned VLA from 52.6% to 6.1%, and increases recovery success from 13% to 80% when refined skills are reused.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2610.21453